{"id":23413,"date":"2026-10-16T06:43:30","date_gmt":"2026-10-16T06:43:30","guid":{"rendered":"https:\/\/www.sphinx-solution.com\/blog\/?p=23413"},"modified":"2026-10-16T06:43:30","modified_gmt":"2026-10-16T06:43:30","slug":"context-engineering-guide","status":"publish","type":"post","link":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/","title":{"rendered":"Context Engineering: A Detailed Guide"},"content":{"rendered":"<div style=\"text-align: left; background-color: #ebf5ff; padding: 20px 25px; font-size: 16px; border-radius: 16px; border: 1px solid #b6dcff; width: 100%; box-sizing: border-box;\">\n<p style=\"font-size: 16px; margin: 0 0 15px 0;\"><strong>Key Takeaways<\/strong><\/p>\n<ul class=\"key-takeaways\" style=\"margin: 0; padding-left: 20px; line-height: 1.7;\">\n<li><span class=\"TextRun SCXW261827081 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW261827081 BCX8\">Context engineering<\/span><\/span><span class=\"TextRun SCXW261827081 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW261827081 BCX8\">supports<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\"> AI models deliver <\/span><span class=\"NormalTextRun SCXW261827081 BCX8\">suitable<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\">, reliable results by\u00a0<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\">offering<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\"> the right information, instructions, memory, and tools for e<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\">very<\/span><span class=\"NormalTextRun SCXW261827081 BCX8\"> task.<\/span><\/span><\/li>\n<li><span class=\"TextRun SCXW156784351 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW156784351 BCX8\">Context engineering vs. prompt engineering<\/span><\/span><span class=\"TextRun SCXW156784351 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW156784351 BCX8\"> comes down to scope: <\/span><\/span><span class=\"TextRun SCXW156784351 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW156784351 BCX8\">prompt engineering<\/span><\/span><span class=\"TextRun SCXW156784351 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW156784351 BCX8\"> focuses on instructions, while context engineering\u00a0<\/span><span class=\"NormalTextRun SCXW156784351 BCX8\">controls<\/span><span class=\"NormalTextRun SCXW156784351 BCX8\"> the broader information the model needs <\/span><span class=\"NormalTextRun SCXW156784351 BCX8\">for effective working<\/span><span class=\"NormalTextRun SCXW156784351 BCX8\">.<\/span><\/span><\/li>\n<li><span class=\"TextRun SCXW163729577 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW163729577 BCX8\">RAG, fine-tuning, and context window management<\/span><\/span><span class=\"TextRun SCXW163729577 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW163729577 BCX8\"> serve different purposes. RAG retrieves\u00a0<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\">suitable<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\"> information, fine-tuning\u00a0<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\">catches<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\"> model behaviour, and context management keeps\u00a0<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\">required<\/span><span class=\"NormalTextRun SCXW163729577 BCX8\"> details accessible without unnecessary clutter.<\/span><\/span><\/li>\n<li><span class=\"TextRun SCXW2049807 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW2049807 BCX8\">Context engineering best practices<\/span><\/span><span class=\"TextRun SCXW2049807 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW2049807 BCX8\">cover<\/span><span class=\"NormalTextRun SCXW2049807 BCX8\"> selecting\u00a0<\/span><span class=\"NormalTextRun SCXW2049807 BCX8\">suited<\/span><span class=\"NormalTextRun SCXW2049807 BCX8\"> information, compressing lengthy content, <\/span><span class=\"NormalTextRun SCXW2049807 BCX8\">maintaining<\/span><span class=\"NormalTextRun SCXW2049807 BCX8\"> useful memory, and connecting AI models to appropriate tools <\/span><span class=\"NormalTextRun SCXW2049807 BCX8\">via <\/span><span class=\"NormalTextRun SCXW2049807 BCX8\">methods<\/span><span class=\"NormalTextRun SCXW2049807 BCX8\"> like <\/span><span class=\"NormalTextRun SCXW2049807 BCX8\">the <\/span><\/span><span class=\"TextRun SCXW2049807 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW2049807 BCX8\">Model Context Protocol (MCP)<\/span><\/span><span class=\"TextRun SCXW2049807 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW2049807 BCX8\">.<\/span><\/span><\/li>\n<li><span class=\"TextRun SCXW196767777 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW196767777 BCX8\">Security and continuous evaluation<\/span><\/span><span class=\"TextRun SCXW196767777 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW196767777 BCX8\"> are essential for dependable AI systems. Validating sources, controlling tool access, and testing accuracy, consistency, cost, and performance help<\/span><span class=\"NormalTextRun SCXW196767777 BCX8\"> in<\/span><span class=\"NormalTextRun SCXW196767777 BCX8\"> reduc<\/span><span class=\"NormalTextRun SCXW196767777 BCX8\">ing<\/span><span class=\"NormalTextRun SCXW196767777 BCX8\"> risks and improve results over time.<\/span><\/span><\/li>\n<\/ul>\n<\/div>\n<p><span data-contrast=\"auto\">For years, we focused heavily on how to think and word a single prompt to get the perfect reply from a Large Language Model (LLM). But as AI is\u00a0evolving, from simple chatbots to tough, autonomous agents able to\u00a0organise multiple-step workflows, just a single good prompt isn&#8217;t enough. One can say, if 2023 was the year\u00a0for prompt engineering, 2026 is officially the\u00a0year of <\/span><b><span data-contrast=\"auto\">Context Engineering<\/span><\/b><span data-contrast=\"auto\">.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Building an AI agent to track customer accounts, remember unresolved support tickets, and maintain a coherent conversation for weeks needs something much bigger. The world of <\/span><b><span data-contrast=\"auto\">Context Engineering <\/span><\/b><span data-contrast=\"auto\">is all about <\/span><span data-contrast=\"auto\">the art and science of dynamically managing an AI\u2019s active memory workspace to keep it accurate, focused, and efficient.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<h2 id=\"what-is-context-engineering\"><strong><span class=\"TextRun SCXW260539666 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW260539666 BCX8\" data-ccp-parastyle=\"heading 2\">What is Context Engineering?<\/span><\/span><span class=\"EOP Selected SCXW260539666 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h2>\n<p><span data-contrast=\"none\">Context engineering is the practice of designing systems that select, organize, and provide the right information, tools, and relevant organizational knowledge to an AI agent. It helps ensure the agent has the context it needs to perform tasks effectively within its context window, which determines how much information the model can process at a time. This information is measured in tokens, the smaller units of text or data processed by the AI model.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">RAG<\/span><\/b><span data-contrast=\"auto\"> pulls in helpful info from outside sources and feeds it into the model so it can answer better. The <\/span><b><span data-contrast=\"auto\">Model Context Protocol<\/span><\/b><span data-contrast=\"auto\">\u00a0allows AI apps link to outside tools, data, and services.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">These techniques always keep the AI model clear on what to do, which data to use, and which tools to pull from. It&#8217;s not\u00a0only about\u00a0collecting more info, but also about shaping what&#8217;s there, so it&#8217;s clear, tidy, and helps with the task.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Now, let us understand the importance of evolution of Context Engineering\u00a0here.<\/span><\/p>\n<h2 id=\"the-emergence-of-context-engineering\"><strong><span class=\"TextRun SCXW102393368 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW102393368 BCX8\" data-ccp-parastyle=\"heading 2\">The Emergence of Context Engineering<\/span><\/span><span class=\"EOP Selected SCXW102393368 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">Context engineering emerged as AI\u00a0developed from simple, single-turn interactions to\u00a0tedious workflows\u00a0including agents, external tools, and multiple decisions. While prompt engineering works well for tasks\u00a0like\u00a0writing emails or summarising text, it becomes less sufficient when AI\u00a0should retrieve information, remember\u00a0past actions, and adapt to\u00a0variable conditions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Modern AI agents work via\u00a0continuous cycles of reasoning, action, and feedback. They can query databases, call APIs, and execute code. <\/span><b><span data-contrast=\"auto\">Context engineering<\/span><\/b><span data-contrast=\"auto\"> manages the resulting information, keeping\u00a0suitable instructions, tool outputs, and task history available without\u00a0burdening the model.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Context overload<\/span><\/b><span data-contrast=\"auto\"> is other challenge. Too much irrelevant information can\u00a0supress important details, increase costs, and affect the quality of response. Context engineering\u00a0shows this through retrieval, filtering, summarisation, and careful context management.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Ultimately, <\/span><b><span data-contrast=\"auto\">context engineering<\/span><\/b><span data-contrast=\"auto\">\u00a0supports build reliable AI applications by information management, connecting tools, and sustaining\u00a0appropriate context\u00a0across a task. This encourages businesses to\u00a0build AI systems handling real-world workflows more efficiently with\u00a0required consistency.<\/span><\/p>\n<h2 id=\"why-context-engineering-matters\"><strong><span class=\"TextRun SCXW255646764 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW255646764 BCX8\" data-ccp-parastyle=\"heading 2\">Why Context Engineering Matters?<\/span><\/span><span class=\"EOP Selected SCXW255646764 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h2>\n<p><b><span data-contrast=\"none\">Context engineering<\/span><\/b><span data-contrast=\"none\">&#8216;s like <\/span><span data-contrast=\"auto\">cleaning<\/span><span data-contrast=\"none\"> up <\/span><span data-contrast=\"auto\">your cluttered desk<\/span><span data-contrast=\"none\"> before you <\/span><span data-contrast=\"auto\">begin.<\/span><span data-contrast=\"none\"> Prompt engineering is about <\/span><span data-contrast=\"auto\">phrasing<\/span><span data-contrast=\"none\"> the <\/span><span data-contrast=\"auto\">question just right.<\/span><span data-contrast=\"none\"> Context engineering <\/span><span data-contrast=\"auto\">clears up<\/span><span data-contrast=\"none\"> the surrounding <\/span><span data-contrast=\"auto\">details<\/span><span data-contrast=\"none\"> so the AI can <\/span><span data-contrast=\"auto\">zero in on<\/span><span data-contrast=\"none\"> the answer.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559737&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"none\">That&#8217;s what lets an AI system stay perfectly accurate, keep results steady, cut costs hard, and never lose track of context. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;201341983&quot;:0,&quot;335551550&quot;:1,&quot;335551620&quot;:1,&quot;335557856&quot;:16777215,&quot;335559738&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:276}\">\u00a0<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;201341983&quot;:0,&quot;335551550&quot;:1,&quot;335551620&quot;:1,&quot;335557856&quot;:16777215,&quot;335559738&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:276}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Properly engineered context directly dictates an AI system&#8217;s viability across four critical pillars: accuracy, consistency, cost, and context rot. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;201341983&quot;:0,&quot;335551550&quot;:1,&quot;335551620&quot;:1,&quot;335557856&quot;:16777215,&quot;335559738&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:276}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">The breakdown of these four\u00a0pillars is as follows:<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:180,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"26\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Accuracy:<\/span><\/b><span data-contrast=\"auto\">\u00a0Eliminates the &#8220;lost in the middle&#8221; condition. By refining noise and placing key data in\u00a0preferred zones, it\u00a0limits hallucinations and forces realistic grounding.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:180}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"26\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Consistency:<\/span><\/b><span data-contrast=\"auto\">\u00a0Set ups model behaviour. Using standard schemas and few-shot examples\u00a0making it sure for identical formatting and tone across millions of different prompts.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:180}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"26\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Cost:<\/span><\/b><span data-contrast=\"auto\">\u00a0looking down API bills and latency. Techniques i.e. Semantic pruning and context caching\u00a0minimise token payloads, making financially\u00a0sound production systems.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:180}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"26\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Context Rot:<\/span><\/b><span data-contrast=\"auto\">\u00a0Context rot avoids\u00a0conversational decay. Dynamic memory management purges stale history while\u00a0saving core data, keeping the AI sharp during long sessions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/li>\n<\/ul>\n<h2 id=\"context-engineering-vs-prompt-engineering-rag-and-fine-tuning\"><strong><span class=\"TextRun SCXW255838006 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW255838006 BCX8\" data-ccp-parastyle=\"heading 2\">Context Engineering vs Prompt Engineering, RAG and Fine-Tuning<\/span><\/span><span class=\"EOP Selected SCXW255838006 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h2>\n<p><span class=\"TextRun SCXW147394126 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW147394126 BCX8\">Selecting the\u00a0<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">perfect<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> approach to\u00a0<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">boost<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> AI responses depends on what the task <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">expects<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">. <\/span><\/span><span class=\"TextRun SCXW147394126 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW147394126 BCX8\">Context engineering vs. prompt engineering, RAG, and fine-tuning<\/span><\/span><span class=\"TextRun SCXW147394126 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW147394126 BCX8\"> is not\u00a0<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">just<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> a comparison of\u00a0techniques, as each <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">one of them <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">addresses a different\u00a0<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">section<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> of the problem. Understanding how they function<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> or work<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">, where they <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">vary<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">, and when to use each <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">this <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">helps <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">to develop<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">\u00a0AI systems<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">, those are<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> delivering more <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">suitable<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">, <\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">accur<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">at<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">e<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">,<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> and<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\"> c<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">on<\/span><span class=\"NormalTextRun SCXW147394126 BCX8\">sistent results.<\/span><\/span><span class=\"EOP Selected SCXW147394126 BCX8\" data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<table data-tablestyle=\"MsoTableGridLight\" data-tablelook=\"1696\" aria-rowcount=\"6\" aria-colcount=\"5\">\n<tbody>\n<tr aria-rowindex=\"1\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Point of Comparison<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Context Engineering<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Prompt Engineering<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">RAG<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Fine-Tuning<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"2\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Primary Goal<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Give the AI the right info, tools, and memory at the right moment<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Get better answers by writing clearer instructions<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Make answers more accurate by pulling in outside information<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Teach the AI a specific skill or style<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"3\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Scope<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">The whole setup around the AI<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Just one question or instruction<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Just the step where information is looked up<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">The AI model itself<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"4\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Flexibility<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Very flexible, adjusts to each task and user<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Fairly flexible, easy to edit but limited to one prompt<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Very flexible, you can swap documents anytime<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Not very flexible, any change means retraining<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"5\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Cost &amp; Effort<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Moderate, needs some planning and setup<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Low, quick and easy to try<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Moderate, needs a system to store and search documents<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">High, needs lots of data, computing power, and expertise<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"6\">\n<td data-celllook=\"0\"><b><span data-contrast=\"auto\">Best For<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Complex AI tools that handle many steps<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Simple tasks and quick tests<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Answering questions from large or changing documents<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Giving the AI a special tone, style, or expertise<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h2 id=\"how-context-engineering-works\"><strong><span class=\"EOP Selected SCXW102393368 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><span class=\"TextRun SCXW107312744 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW107312744 BCX8\" data-ccp-parastyle=\"heading 2\">How\u00a0Context Engineering Works?<\/span><\/span><span class=\"EOP Selected SCXW107312744 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">The current scenario of AI applications is going beyond simple question-and-answer interactions. Core\u00a0requirements of AI agents built on large language models (LLMs)\u00a0comprises of\u00a0user requests processing, retrieving enterprise knowledge,\u00a0utilisation of external tools, managing context memory, and\u00a0finishing up multi-turn tasks. <\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">This is where context engineering is useful. Rather than iterating on the prompt, engineers build out a full information flow that describes what data, instructions, history, and tool outputs an AI model sees before it produces a response. <\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">A good context system lets AI applications filter out irrelevant information, increase accuracy, manage token usage and produce more reliable results in production environments.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<h3 id=\"the-context-pipeline\" aria-level=\"3\"><strong>The Context Pipeline\u00a0<\/strong><\/h3>\n<p><span data-contrast=\"auto\">The context pipeline elaborates how information\u00a0shifts from a user&#8217;s request to the AI-generated response. It\u00a0also supports developers to understand where information is accumulated, how it is filtered \/ refined, and how it is\u00a0transferred to a large language model (LLM).<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">A context pipeline follows these stages:<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Receive Request:<\/span><\/b><span data-contrast=\"auto\"> Know the user first. The system\u00a0identifies the question, the reason for it, and the\u00a0style the answer has to be given. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Collect relevant information:<\/span><\/b><span data-contrast=\"auto\"> The request is\u00a0registered\u00a0along with the system instructions. That is often sufficient, but when it is not, the system pulls up what else is needed, be it a document, a database record, a previous message, or something in memory. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Context setting:<\/span><\/b><span data-contrast=\"auto\"> This is the cleanup stage. Irrelevant details are stripped away, redundancy is removed, and the data is checked for currency wherever possible. The rest is used for a reasonable order.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Manage the context window: <\/span><\/b><span data-contrast=\"auto\">It estimates the token usage and reserves space for instructions, retrieved information, conversation history, tool definitions, and the model\u2019s expected response.\u00a0 <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Generate or act: <\/span><\/b><span data-contrast=\"auto\">The model either generates an answer or asks for an appropriate tool to either fetch more data or perform an authorised action.\u00a0 <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Match the response with the relevant requirements and apply the appropriate safeguards. Return the answer. If the task needs more information, the process repeats.<\/span><\/p>\n<h3 id=\"context-window-limits-and-token-budgeting\"><strong><span class=\"TextRun SCXW117291986 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW117291986 BCX8\" data-ccp-parastyle=\"heading 3\">Context window limits and token budgeting<\/span><\/span><span class=\"EOP Selected SCXW117291986 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">A model may handle\u00a0a large amount of text at once. Whatever goes in,\u00a0with whatever comes back out,\u00a0should fit inside its context window, and that space is\u00a0counted in tokens. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Token budgeting simply means planning how that space gets shared out. Some goes to instructions, some to the user\u2019s question, some to retrieved documents, tool descriptions and past messages, and some has to be kept free for the reply.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Think of a support agent three weeks into a customer\u2019s issue. Pasting the whole thread into every request is wasteful. It works better to keep a running summary and fetch an older message only when the customer refers to it. That cuts token use, and the model isn\u2019t wading through details that no longer matter.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Effective context management prioritises accurate, useful information while\u00a0saving essential instructions and security constraints.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<h2 id=\"core-components-of-context-engineering\"><strong><span class=\"TextRun SCXW160670307 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW160670307 BCX8\" data-ccp-parastyle=\"heading 2\">Core Components of Context Engineering<\/span><\/span><span class=\"EOP Selected SCXW160670307 BCX8\" data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559737&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/strong><\/h2>\n<p><img decoding=\"async\" class=\"alignnone wp-image-23415 size-full\" src=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Core-Components-of-Context-Engineering_.webp\" alt=\"Core Components of Context Engineering\" width=\"700\" height=\"343\" srcset=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Core-Components-of-Context-Engineering_.webp 700w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Core-Components-of-Context-Engineering_-300x147.webp 300w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Core-Components-of-Context-Engineering_-390x191.webp 390w\" sizes=\"(max-width: 700px) 100vw, 700px\" \/><\/p>\n<p><span data-contrast=\"auto\">Core components of context engineering<\/span><span data-contrast=\"auto\">\u00a0emphasis<\/span><span data-contrast=\"auto\"> on optimizing how information is selected, streamlined, and delivered to large language models (LLMs) to maximize reply to accuracy<\/span><span data-contrast=\"auto\">. This discipline\u00a0includes <\/span><span data-contrast=\"auto\">dynamic context retrieval<\/span><span data-contrast=\"auto\"> to\u00a0extract only the most\u00a0suitable data, alongside <\/span><span data-contrast=\"auto\">prompt compression<\/span><span data-contrast=\"auto\"> to minimize token usage and costs. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559737&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">It also encompasses <\/span><span data-contrast=\"auto\">metadata enrichment<\/span><span data-contrast=\"auto\"> to\u00a0offer critical situational tracking and <\/span><span data-contrast=\"auto\">structural formatting<\/span><span data-contrast=\"auto\"> to\u00a0make sure that the model parses information efficiently. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559737&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">These elements change raw data into a streamlined, high-utility input which directly improves an AI&#8217;s reasoning capabilities.<\/span><\/p>\n<p><span data-contrast=\"auto\">Let us roll them out here:<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335559737&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p data-pm-slice=\"1 1 []\"><strong>\u2022 Prompting Techniques<\/strong><\/p>\n<p>Classic prompting is related to pattern recognition. It involves providing examples, specifying the desired format, and defining clear guidelines for the model to follow. Few-shot prompting works particularly well for structured tasks.<\/p>\n<p>Advanced prompting techniques are even more interesting. For example, Chain-of-Thought (CoT) prompting encourages a model to work through complex problems in a structured manner.<\/p>\n<p><strong>\u2022 Query Augmentation<\/strong><\/p>\n<p>Users are often reluctant to type long, detailed queries repeatedly. Query augmentation addresses this challenge by improving and expanding user queries through the following techniques:<\/p>\n<ul data-spread=\"false\">\n<li><strong>Query Rewriting:<\/strong> An LLM takes a vague or unclear question and reformulates it to make the user&#8217;s intent clearer.<\/li>\n<li><strong>Query Expansion:<\/strong> Adds related terms, synonyms, and relevant keywords to broaden the search and retrieve more useful results.<\/li>\n<li><strong>Query Decomposition:<\/strong> Breaks a complex question into smaller sub-queries that can be addressed independently.<\/li>\n<li><strong>Query Agents:<\/strong> Use an agent to dynamically determine how to reformulate or refine a query based on initial results.<\/li>\n<\/ul>\n<p><strong>\u2022 Long-Term Memory<\/strong><\/p>\n<p>An AI agent may have in-depth conversations with users who share preferences, contextual information, and personal history. However, this information can be lost when a session ends.<\/p>\n<p>Long-term memory addresses this limitation by storing relevant information in external systems, allowing agents to retrieve useful details across multiple sessions.<\/p>\n<ul data-spread=\"false\">\n<li><strong>Vector Databases:<\/strong> Store embeddings of past interactions and enable semantic searches to retrieve relevant information based on meaning.<\/li>\n<li><strong>Graph Databases:<\/strong> Store entities and their relationships to represent connections between users, conversations, and contextual information.<\/li>\n<\/ul>\n<p>Open-source tools make these capabilities more accessible, eliminating the need to build every component from scratch.<\/p>\n<p><strong>\u2022 Short-Term Memory<\/strong><\/p>\n<p>Short-term memory refers to the conversation history and contextual information an AI agent uses during an ongoing interaction. Although this capability is often available by default, it can be poorly managed.<\/p>\n<p>Common challenges include:<\/p>\n<ul data-spread=\"false\">\n<li><strong>Context Overloading:<\/strong> Including too much information in the context window, potentially reducing the model&#8217;s effectiveness.<\/li>\n<li><strong>Insufficient Context:<\/strong> Failing to provide enough relevant information for the model to generate accurate responses.<\/li>\n<li><strong>Poor Context Ordering:<\/strong> Placing important information where the model may be less likely to use it effectively.<\/li>\n<li><strong>Lack of Summarization:<\/strong> Failing to summarize long conversations, resulting in unnecessary context consumption and the loss of important details.<\/li>\n<\/ul>\n<p><strong>\u2022 Comparison: Short-Term Memory vs. Long-Term Memory<\/strong><\/p>\n<p>Memory enables AI agents to track conversations, retain useful information, and provide more relevant responses. Short-term memory focuses on the current interaction, while long-term memory preserves important information across sessions.<\/p>\n<table data-tablestyle=\"MsoTable15Grid5DarkAccent1\" data-tablelook=\"1696\" aria-rowcount=\"6\" aria-colcount=\"3\">\n<tbody>\n<tr aria-rowindex=\"1\">\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">Aspect<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">Short-term memory<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">Long-term memory<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"2\">\n<td data-celllook=\"4096\"><b><span data-contrast=\"none\">Purpose<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Maintains context during the current conversation or task.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Retains useful information across conversations or sessions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"3\">\n<td data-celllook=\"4096\"><b><span data-contrast=\"none\">Stores<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Recent messages, current goals, intermediate results, and decisions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">User preferences, important facts, and\u00a0suitable past interactions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"4\">\n<td data-celllook=\"4096\"><b><span data-contrast=\"none\">Duration<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Usually session-based; may be summarised or cleared.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Persists in external storage until updated or deleted.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"5\">\n<td data-celllook=\"4096\"><b><span data-contrast=\"none\">How it works<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Uses conversation history or active task state.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Retrieves stored information when it becomes relevant.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"6\">\n<td data-celllook=\"4352\"><b><span data-contrast=\"none\">Example<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Remembering the destination and budget while planning a trip.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-contrast=\"auto\">Remembering the user&#8217;s general travel preferences for future trips.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p><strong>\u2022 Retrieval<\/strong><\/p>\n<p>Many teams consider this as RAG, but this is very stiff. RAG is just a pattern.<\/p>\n<p>The real challenge lies in connecting AI technology with your business data.<\/p>\n<p>Those sources are scattered in the form of docs, Wikipedia, databases, SaaS tools such as Notion and Google Drive, APIs, and code repositories.<\/p>\n<p>The retrieval pipeline comes up across three layers:<\/p>\n<ul>\n<li>Pre-Retrieval<\/li>\n<li>Retrieval<\/li>\n<li>Augmentation<\/li>\n<\/ul>\n<p>Open-source tools handle it all without needing custom connectors for every data source. Just sync your knowledge bases and access Notion, Google Drive, databases, and other tools from one place.<\/p>\n<p><strong>\u2022 Tools and Agents<\/strong><\/p>\n<p>A tool extends what the model can do because, without it, the model is hampered with just what\u2019s in its weights and context window. Also, an agent decides when and how to use these tools.<\/p>\n<h2 id=\"four-core-strategies-for-context-engineering\"><strong><span class=\"TextRun SCXW15314742 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW15314742 BCX8\" data-ccp-parastyle=\"heading 2\">Four Core Strategies for Context Engineering<\/span><\/span><span class=\"EOP Selected SCXW15314742 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h2>\n<p><img decoding=\"async\" class=\"alignnone wp-image-23416 size-full\" src=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Four-Core-Strategies-for-Effective-Context-Engineering_.webp\" alt=\"Four Core Strategies for Effective Context Engineering\" width=\"700\" height=\"345\" srcset=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Four-Core-Strategies-for-Effective-Context-Engineering_.webp 700w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Four-Core-Strategies-for-Effective-Context-Engineering_-300x148.webp 300w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Four-Core-Strategies-for-Effective-Context-Engineering_-390x192.webp 390w\" sizes=\"(max-width: 700px) 100vw, 700px\" \/><\/p>\n<p><span data-contrast=\"auto\">An AI model\u00a0may lose track of important details when a task\u00a0has much lengthier conversations,\u00a0several documents, or many\u00a0steps. <\/span><b><span data-contrast=\"auto\">Context engineering<\/span><\/b><span data-contrast=\"auto\">\u00a0supports in managing this\u00a0situation by deciding what information to save, retrieve, condense, and keep separate. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Four core strategies i.e. <\/span><span data-contrast=\"auto\">Write, Select, Compress, and Isolate.<\/span><span data-contrast=\"auto\"> Together, they help AI agents to use their context windows more effectively without any loss of the information\u00a0required to\u00a0do a task.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><strong>\u2022 Write: Save Information<\/strong><\/p>\n<p>AI agent jots down thoughts, plans, and discoveries while working through problems. Keeps facts and preferences stored away to use later. The AI keeps facts and preferences around for later use. Storing information outside the context window also leaves more room for the details needed now.<\/p>\n<p><strong>\u2022 Select: What Exactly Require<\/strong><\/p>\n<p>Retrieves only the appropriate data, saved memories, and tool definitions needed for the existing task. It uses agentic Retrieval-Augmented Generation (RAG), semantic similarity search, and just-in-time tool to reduce unnecessary token usage and keep the AI focused on the current task.<\/p>\n<p><strong>\u2022 Compress: Keep Useful Tokens<\/strong><\/p>\n<p>Long interactions produce far more text than a model can use well. Summarisation condenses previous turns, and trimming trims out older or repeated messages. A weak summary is risky, because any detail it skips is missing from each further step.<\/p>\n<p><strong>\u2022 Isolate: Separate Non-related Work<\/strong><\/p>\n<p>Dividing a job among sub-agents gives each one a dedicated window and a narrow focus. Sandboxed environments do something similar for bulky outputs. The price is higher.<\/p>\n<h2 id=\"tools-and-mcp-in-context-engineering\"><strong><span class=\"TextRun SCXW219337295 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW219337295 BCX8\" data-ccp-parastyle=\"heading 2\">Tools and MCP in Context Engineering<\/span><\/span><span class=\"EOP Selected SCXW219337295 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">AI models\u00a0mainly need capabilities or informational details which are unavailable in the conversation itself. Tools enable an AI application to retrieve documents, querying databases, check records,\u00a0conduct calculations, or execute other\u00a0set operations. Context engineering\u00a0detects which tools are ready for use, when they\u00a0must be used, and how their results become integral part of the model&#8217;s working context. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Refer a table for better understanding here<\/span><\/p>\n<table data-tablestyle=\"MsoTable15Grid5DarkAccent1\" data-tablelook=\"1696\" aria-rowcount=\"9\" aria-colcount=\"3\">\n<tbody>\n<tr aria-rowindex=\"1\">\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">Category<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">MCP Tool \/ Integration<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"4113\"><b><span data-contrast=\"none\">Description &amp; Capabilities<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"2\">\n<td colspan=\"1\" rowspan=\"3\" data-celllook=\"4096\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"none\">Development &amp; Code Repositories<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">GitHub<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Handles repository management actions and tracking (e.g., create_issue, get_issue).<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"3\">\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Filesystem<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Provides local directory sandboxing for secure file reading, writing, and editing.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"4\">\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">SQL Server \/ Databases<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Enables conversational database querying and schema inspection (e.g., list_databases, get_schema).<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"5\">\n<td data-celllook=\"4096\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"none\">Productivity &amp; Communication<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Google Calendar<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Handles schedule management through functions like get_calendar, create_event, and update_event.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"6\">\n<td colspan=\"1\" rowspan=\"3\" data-celllook=\"4096\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"none\">Web Research &amp; Automation<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Fetch<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Retrieves and converts external web content into token-efficient formats for LLMs.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"7\">\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Firecrawl<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Executes deep web scraping and site-mapping tasks (e.g., start crawl, start_batch_scrape).<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"8\">\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Playwright<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Powers end-to-end browser automation and web application testing.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<tr aria-rowindex=\"9\">\n<td data-celllook=\"4352\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"none\">Reasoning &amp; Problem-Solving<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Sequential Thinking<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<td data-celllook=\"0\"><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Guides the model through a dynamic, step-by-step reflection process for architectural debugging and complex problem-solving.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:2,&quot;335551620&quot;:2,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<h3 id=\"designing-clear-tool-descriptions\"><strong><span class=\"TextRun SCXW231743630 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW231743630 BCX8\" data-ccp-parastyle=\"heading 3\">Designing Clear Tool Descriptions<\/span><\/span><span class=\"EOP Selected SCXW231743630 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">Building good tools means getting the basics right<\/span><span data-contrast=\"auto\">. Name it clearly.<\/span><span data-contrast=\"auto\"> Know what it does. Define your inputs and outputs. Make them predictable. Leave balance checks need different tools than HR policy lookups. Employee support agents use both. When descriptions are solid, models pick the right tool. You waste fewer tokens. Functions don&#8217;t overlap or confuse things. <\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Security primarily depends on application-level input validation and access controls.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<h3 id=\"how-the-model-context-protocol-mcp-helps\"><strong><span class=\"TextRun SCXW221810068 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW221810068 BCX8\" data-ccp-parastyle=\"heading 3\">How the Model Context Protocol (MCP) Helps<\/span><\/span><span class=\"EOP Selected SCXW221810068 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">MCP, the Model Context Protocol, standardizes how AI applications\u00a0communicate with external tools and data sources. AI apps can discover what capabilities MCP servers expose. Fetch company documents. Access employee records you&#8217;re authorized to see. Invoke those capabilities when needed.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">This supports <\/span><b><span data-contrast=\"auto\">context engineering<\/span><\/b><span data-contrast=\"auto\"> by making\u00a0suitable external information available to the model. Anyhow, MCP does not independently manage the entire context window or guarantee accurate responses.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<h3 id=\"combining-tools-retrieval-and-context-management\"><strong><span class=\"TextRun SCXW133218393 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW133218393 BCX8\" data-ccp-parastyle=\"heading 3\">Combining Tools, Retrieval, and Context Management<\/span><\/span><span class=\"EOP Selected SCXW133218393 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">Effective <\/span><b><span data-contrast=\"auto\">context engineering<\/span><\/b><span data-contrast=\"auto\">\u00a0aligns tools, retrieval-augmented generation (RAG), and memory to\u00a0deliver relevant information when required. RAG retrieves external information, tools access data or perform defined actions, and memory retains useful details across conversations. Together, these components help AI systems handle multistep tasks while\u00a0organising token usage and validating results.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Unlike <\/span><b><span data-contrast=\"auto\">context engineering vs prompt engineering<\/span><\/b><span data-contrast=\"auto\">, which compares two approaches to guide AI models, this section entrusts on how tools, retrieval, and memory work together. In a broader <\/span><b><span data-contrast=\"auto\">context model in software engineering<\/span><\/b><span data-contrast=\"auto\">, developers also define the entities, relationships, and environmental information- a system needs to operate correctly.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<h2 id=\"real-world-examples-of-context-engineering\"><strong><span class=\"TextRun SCXW247317803 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW247317803 BCX8\" data-ccp-parastyle=\"heading 2\">Real-World Examples of Context Engineering<\/span><\/span><span class=\"EOP Selected SCXW247317803 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h2>\n<p><img decoding=\"async\" class=\"alignnone wp-image-23417 size-full\" src=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Real-World-Examples-of-Context-Engineering_.webp\" alt=\"Real-World Examples of Context Engineering\" width=\"700\" height=\"351\" srcset=\"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Real-World-Examples-of-Context-Engineering_.webp 700w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Real-World-Examples-of-Context-Engineering_-300x150.webp 300w, https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Real-World-Examples-of-Context-Engineering_-390x196.webp 390w\" sizes=\"(max-width: 700px) 100vw, 700px\" \/><\/p>\n<p><span class=\"TextRun SCXW261292251 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW261292251 BCX8\">Context engineering<\/span><\/span><span class=\"TextRun SCXW261292251 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"auto\"><span class=\"NormalTextRun SCXW261292251 BCX8\"> helps AI deliver useful and <\/span><span class=\"NormalTextRun SCXW261292251 BCX8\">accurate<\/span><span class=\"NormalTextRun SCXW261292251 BCX8\"> results by supplying the information, tools, and conversation history\u00a0<\/span><span class=\"NormalTextRun SCXW261292251 BCX8\">suitable<\/span><span class=\"NormalTextRun SCXW261292251 BCX8\"> to a particular task. This makes it easier for AI to work through complex, multi-step tasks in areas such as customer support, software development, and research, where a simple prompt may not provide enough information to produce the desired outcome.<\/span><\/span><\/p>\n<p><strong>\u2022 Customer Support Bot<\/strong><\/p>\n<p>Aligns customer history, authorised account details, and support documents to solve queries without repeating troubleshooting steps. RAG gathers suitable information, while tools check ticket status.<\/p>\n<p><strong>\u2022 Coding Assistant<\/strong><\/p>\n<p>Uses source code, project documentation, and error logs to suggest suitable solutions. Repository tools and test results support validating changes against the current codebase.<\/p>\n<p><strong>\u2022 Research Agent<\/strong><\/p>\n<p>Retrieves information from multiple sources, compares findings, and identifies conflicting claims. Context management saves suitable evidence while controlling token usage.<\/p>\n<p><strong>\u2022 Before-and-After Case<\/strong><\/p>\n<p><strong>Before:<\/strong> An AI assistant gives a common answer about employee leave carry-forward without checking company policy.<\/p>\n<p><strong>After:<\/strong> It retrieves the applicable policy and authorised employee details to offer a suitable answer or request clarification.<\/p>\n<p>This example shows how a well-designed context pipeline improves response quality via the most suitable retrieval, verified information, and appropriate access controls.<\/p>\n<h2 id=\"how-to-implement-context-engineering-in-ai-products-and-business-workflows\"><strong><span class=\"TextRun SCXW114949305 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW114949305 BCX8\" data-ccp-parastyle=\"heading 2\">How to Implement Context Engineering in AI Products and Business Workflows<\/span><\/span><span class=\"EOP Selected SCXW114949305 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\">\u00a0<\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">Applying or utilising <\/span><b><span data-contrast=\"auto\">context engineering<\/span><\/b><span data-contrast=\"auto\"> starts with a basic understanding, what an AI application\u00a0required to deliver reliable results. A structured approach\u00a0supports teams to select\u00a0certain information, connect with the right tools, manage token usage, and maintain security as the solution\u00a0progresses from testing to production stage.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Define your task at start:<\/span><\/b><span data-contrast=\"auto\"> What&#8217;s the business problem? What output do you want? How do you know it worked out? <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Identify context:<\/span><\/b><span data-contrast=\"auto\"> What instructions, documents, user data, and chat history does your model need to see? <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Build out your context pipeline: <\/span><\/b><span data-contrast=\"auto\">Wire up retrieval systems, memory, tools like APIs or MCP servers. Use what makes sense. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Manage and secure the data: <\/span><\/b><span data-contrast=\"auto\">Strip out noise. Control who accesses what. Keep token usage inside your context window. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"5\" data-aria-level=\"1\"><span data-contrast=\"auto\">T<\/span><b><span data-contrast=\"auto\">est and evaluate all:<\/span><\/b><span data-contrast=\"auto\">\u00a0Test accuracy, retrieval quality, response consistency, latency, cost, security risks. Run through it all. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"75\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"6\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Deploy in stages:\u00a0<\/span><\/b><span data-contrast=\"auto\">See how it performs in the\u00a0actual world.\u00a0Align your context selection as business requirements shifts. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<p><span data-contrast=\"auto\">Context engineering isn&#8217;t a one-time setup. Treat it as\u00a0continuous work.\u00a0Regular review and improvement\u00a0sustain\u00a0usefulness, safety and matching perfectly with your dynamic business needs.<\/span><\/p>\n<h2 id=\"best-practices-common-mistakes-and-security\"><strong><span class=\"TextRun SCXW77640800 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW77640800 BCX8\" data-ccp-parastyle=\"heading 2\">Best Practices, Common Mistakes and Security<\/span><\/span><span class=\"EOP Selected SCXW77640800 BCX8\" data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:299,&quot;335559739&quot;:299}\">\u00a0<\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">AI models with powerful tools, extensive knowledge, and huge context windows still mess up if they get irrelevant information, outdated data, or contradictory inputs. Context engineering fixes this. Get the right information to the model at the right moment. Keep instructions, data sources, and available actions clearly defined.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Anyhow, making a reliable AI system needs more than selecting useful context. Developers must also manage token usage, validate retrieved information, protect sensitive data, and\u00a0stop malicious instructions from\u00a0affecting the model. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Following\u00a0best practices helps in improvement of response quality, reduce common context failures, and make AI applications safer.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<h3 id=\"context-engineering-best-practices-checklist\"><strong><span class=\"TextRun SCXW69638314 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW69638314 BCX8\" data-ccp-parastyle=\"heading 3\">Context Engineering Best Practices Checklist<\/span><\/span><span class=\"EOP Selected SCXW69638314 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">Effective context engineering depends on selecting relevant information, managing the context window, and maintaining control over data and tools. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Follow\u00a0this checklist when designing \/ improving an AI application.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Define clear objectives:<\/span><\/b><span data-contrast=\"auto\"> Specify the task, desired output, and essential instructions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Retrieve relevant information:<\/span><\/b><span data-contrast=\"auto\"> Use retrieval-augmented generation (RAG) to supply useful, up-to-date information.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Manage the context window:<\/span><\/b><span data-contrast=\"auto\"> Allocate tokens across instructions, retrieved data, conversation history, tool results, and the final response.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Maintain data quality:<\/span><\/b><span data-contrast=\"auto\"> Remove old, outdated information and verify source reliability.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"5\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Design clear tools:<\/span><\/b><span data-contrast=\"auto\"> Define each tool&#8217;s purpose, inputs, outputs, and permissions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"6\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Protect sensitive data:<\/span><\/b><span data-contrast=\"auto\"> Retrieve only the information needed for the task and enforce access controls.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"7\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Prevent prompt injection:<\/span><\/b><span data-contrast=\"auto\"> Treat external content as untrusted and validate tool calls before execution.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"42\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"8\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Test and monitor:<\/span><\/b><span data-contrast=\"auto\">\u00a0Assess response accuracy, retrieval quality, token usage, and security risks.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<p><span data-contrast=\"auto\">Apply\u00a0this checklist when designing \/ improving an AI application:<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><span data-contrast=\"auto\">\u00a0<\/span><b><span data-contrast=\"auto\">Define clear objectives:<\/span><\/b><span data-contrast=\"auto\"> What&#8217;s the task. What output do you want. What instructions matter most.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">\u00a0Retrieve relevant information:<\/span><\/b><span data-contrast=\"auto\"> Use retrieval-augmented generation to supply useful, current information.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><span data-contrast=\"auto\">\u00a0<\/span><b><span data-contrast=\"auto\">Manage the context window: <\/span><\/b><span data-contrast=\"auto\">Split your tokens across instructions, retrieved data, conversation history, tool results, and the final response.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><span data-contrast=\"auto\">\u00a0<\/span><b><span data-contrast=\"auto\">Maintain data quality: <\/span><\/b><span data-contrast=\"auto\">Remove outdated information. Verify sources work.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"5\" data-aria-level=\"1\"><span data-contrast=\"auto\">\u00a0<\/span><b><span data-contrast=\"auto\">Design clear tools:<\/span><\/b><span data-contrast=\"auto\"> Each tool needs a purpose. Specify inputs, outputs, and permissions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"6\" data-aria-level=\"1\"><span data-contrast=\"auto\">\u00a0<\/span><b><span data-contrast=\"auto\">Protect sensitive data: <\/span><\/b><span data-contrast=\"auto\">Only retrieve what you need. Enforce access controls.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"7\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Prevent prompt injection: <\/span><\/b><span data-contrast=\"auto\">Treat external content as untrusted. Validate tool calls before running them.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"76\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"8\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Test and monitor: <\/span><\/b><span data-contrast=\"auto\">Check response accuracy. Look at retrieval quality. Track token usage. Watch for security risks.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<p><span data-contrast=\"auto\">Good context engineering means picking relevant information, managing your context window, and controlling data and tools. Use this when building or improving an AI application.<\/span><\/p>\n<h3 id=\"common-context-engineering-mistakes\"><strong><span class=\"TextRun SCXW125710712 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW125710712 BCX8\" data-ccp-parastyle=\"heading 3\">Common Context Engineering Mistakes<\/span><\/span><span class=\"EOP Selected SCXW125710712 BCX8\" data-ccp-props=\"{&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335559738&quot;:160,&quot;335559739&quot;:80}\"><br \/>\n<\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">Also, a well-designed AI workflow\u00a0is possible for the failure, when its context\u00a0has unreliable information or data,\u00a0too many details, or conflicting instructions. <\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Four common failure modes help developers identify the cause and\u00a0use the right correction.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"44\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Context poisoning:<\/span><\/b><span data-contrast=\"auto\"> Incorrect, old, malicious information\u00a0goes into the context and effects replies. Validate sources, refresh stale records, and treat retrieved content as\u00a0much untrusted.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"44\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Context distraction:<\/span><\/b><span data-contrast=\"auto\">\u00a0Extended conversation history \/ irrelevant documents distract from the\u00a0current task. Summarise lengthy interactions and retrieve suitable information only.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"44\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Context confusion:<\/span><\/b><span data-contrast=\"auto\"> Overlapping tools \/ unnecessary details make it harder for the model to go for\u00a0the right information or action. Streamline the context and clear out tool descriptions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"44\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Context clash:<\/span><\/b><span data-contrast=\"auto\"> Conflicting facts or instructions\u00a0take to inconsistent answers. Check source dates and versions,\u00a0set up source-priority rules, and flag unresolved conflicts\u00a0over just guessing.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<h3 id=\"security-privacy-and-prompt-injection\"><strong><span class=\"TextRun SCXW21532771 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW21532771 BCX8\" data-ccp-parastyle=\"heading 3\">Security, Privacy and Prompt Injection<br \/>\n<\/span><\/span><\/strong><\/h3>\n<p><span data-contrast=\"auto\">Prompt injection<\/span><span data-contrast=\"auto\">\u00a0happens when malicious instructions in user input or external content try to manipulate an AI model or\u00a0carrying out unauthorised actions. Such instructions may\u00a0be there in documents, emails, web pages, or tool outputs, making security a required part of <\/span><b><span data-contrast=\"auto\">context engineering<\/span><\/b><span data-contrast=\"auto\">.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Developers must\u00a0apply or use layered safeguards to lower down these risks:<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"1\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Separate data instructions:<\/span><\/b><span data-contrast=\"auto\">\u00a0Separate\u00a0system instructions from user input and retrieved content. Delimiters\u00a0can help. But they\u00a0don\u2019t guarantee safety.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"2\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Implement access controls:<\/span><\/b><span data-contrast=\"auto\"> Verify identity and permissions in the application or retrieval layer\u00a0instead of relying on the model\u00a0for the access.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"3\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Validate tool calls:<\/span><\/b><span data-contrast=\"auto\"> Check inputs, outputs, and permissions before\u00a0running operations.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"4\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Restrict sensitive actions:<\/span><\/b><span data-contrast=\"auto\">\u00a0Require appropriate authorisation or human approval for high-impact operations,\u00a0like deleting records or changing permissions.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"5\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Restrict sensitive information:<\/span><\/b><span data-contrast=\"auto\"> Retrieve necessary\u00a0personal or confidential data only and protect stored memory, logs.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<ul>\n<li aria-setsize=\"-1\" data-leveltext=\"\uf0b7\" data-font=\"Symbol\" data-listid=\"45\" data-list-defn-props=\"{&quot;335552541&quot;:1,&quot;335559683&quot;:0,&quot;335559684&quot;:-2,&quot;335559685&quot;:720,&quot;335559991&quot;:360,&quot;469769226&quot;:&quot;Symbol&quot;,&quot;469769242&quot;:[8226],&quot;469777803&quot;:&quot;left&quot;,&quot;469777804&quot;:&quot;\uf0b7&quot;,&quot;469777815&quot;:&quot;hybridMultilevel&quot;}\" data-aria-posinset=\"6\" data-aria-level=\"1\"><b><span data-contrast=\"auto\">Monitor and test:<\/span><\/b><span data-contrast=\"auto\"> Test direct and indirect prompt-injection attempts,\u00a0check again security events, and update safeguards with\u00a0varying threats.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/li>\n<\/ul>\n<p><span data-contrast=\"auto\">No single prompt or filter\u00a0guarantees complete protection. Mix of\u00a0access controls, input validation, monitoring, and human\u00a0check offers a strong security approach.<\/span><\/p>\n<h2 id=\"conclusion\"><strong><span class=\"TextRun SCXW227164440 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW227164440 BCX8\" data-ccp-parastyle=\"heading 2\">Conclusion<\/span><\/span><span class=\"EOP Selected SCXW227164440 BCX8\" data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:299,&quot;335559739&quot;:299}\">\u00a0<\/span><\/strong><\/h2>\n<p><span data-contrast=\"auto\">Context engineering is becoming\u00a0more important due to dynamical change in AI technologies from basic prompt and response approach to the automatic workflows and doing multiple tasks at a time with AI agents. A\u00a0great, phrased prompt can\u00a0change a model, but the output quality is entirely depending on the fed data, the tools it can use, and how you\u00a0structure its context (with a task).<\/span><span data-contrast=\"auto\">\u00a0<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">As explained in this guide, successful context engineering combines relevant data, retrieval, memory, token budgeting, and safe tool integration to achieve better accuracy, consistency, and efficiency. <\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Whether you\u2019re building a bot to support customers, an assistant to help with coding, or an AI agent to run business workflows, the goal is the same: give the model the right context, at the right time, with the right safeguards. And that\u2019s what makes the difference between promising AI capability and a reliable solution for the real world.<\/span><\/p>\n<h2 id=\"faqs\"><strong><span class=\"TextRun SCXW180133454 BCX8\" lang=\"EN-GB\" xml:lang=\"EN-GB\" data-contrast=\"none\"><span class=\"NormalTextRun SCXW180133454 BCX8\" data-ccp-parastyle=\"heading 2\">FAQs<\/span><\/span><\/strong><\/h2>\n<p><b><span data-contrast=\"auto\">Is context engineering replacing prompt engineering?<\/span><\/b><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Context engineering is\u00a0shifting from a\u00a0trending word to a core discipline\u00a0which is superseding and expanding upon standalone prompt engineering in production AI systems.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Is RAG part of context engineering?<\/span><\/b><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">Retrieval-Augmented Generation (RAG) is a\u00a0certain technique, which is a definite part of context engineering.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">How does MCP relate to context engineering?<\/span><\/b><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">The <\/span><b><span data-contrast=\"auto\">Model Context Protocol (MCP)<\/span><\/b><span data-contrast=\"auto\">\u00a0works<\/span><span data-contrast=\"auto\"> as the standard bridge and infrastructure layer for context engineering. <\/span><span data-contrast=\"auto\">MCP\u00a0offers the universal connector and interface contract\u00a0to make dynamic context retrieval scalable.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Do you need context engineering for simple chatbots?<\/span><\/b><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">No. Context engineering is not required for simple chatbots. Simple chatbots are based on question-and-answer format or query and response technique.<\/span><span data-ccp-props=\"{}\">\u00a0<\/span><\/p>\n<p aria-level=\"3\"><b><span data-contrast=\"auto\">What is context poisoning, and how does it happen?<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:281,&quot;335559739&quot;:281}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"none\">Context poisoning<\/span><\/b><span data-contrast=\"none\"> is when an AI system\u2019s working context is compromised by incorrect, misleading or malicious information and influences its responses. Later decisions can be affected if information isn&#8217;t properly controlled. This requires verification of the source, reliable information retrieval, and prudent memory management.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;201341983&quot;:0,&quot;335551550&quot;:1,&quot;335551620&quot;:1,&quot;335557856&quot;:16777215,&quot;335559738&quot;:0,&quot;335559739&quot;:0,&quot;335559740&quot;:276}\">\u00a0<\/span><\/p>\n<p aria-level=\"3\"><b><span data-contrast=\"auto\">What is the enterprise context gap in AI systems?<\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;134245418&quot;:true,&quot;134245529&quot;:true,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:281,&quot;335559739&quot;:281}\">\u00a0<\/span><\/p>\n<p><span data-contrast=\"auto\">The <\/span><span data-contrast=\"auto\">enterprise context gap<\/span><span data-contrast=\"auto\"> is the difference between an AI model\u2019s general knowledge and the certain information needing for a business. Connecting it to\u00a0certain company policies, documentation, and authorised data\u00a0supports it produce more useful, business-driven responses.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559738&quot;:240,&quot;335559739&quot;:240}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">How does context engineering support multimodal AI? <\/span><\/b><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559685&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n<p><b><span data-contrast=\"auto\">Context engineering<\/span><\/b><span data-contrast=\"auto\"> assists multimodal AI systems in better understanding inputs across text, image, audio and video. It gives multimodal AI models the flexibility to pull context and make links between data in different formats to allow for more effective responses.<\/span><span data-ccp-props=\"{&quot;134233117&quot;:false,&quot;134233118&quot;:false,&quot;335551550&quot;:0,&quot;335551620&quot;:0,&quot;335559685&quot;:0,&quot;335559738&quot;:0,&quot;335559739&quot;:0}\">\u00a0<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"Key Takeaways Context engineeringsupports AI models deliver suitable, reliable results by\u00a0offering the right information, instructions, memory, and tools for every task. Context engineering vs. prompt engineering comes down to scope:&hellip;\n","protected":false},"author":21,"featured_media":23421,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"inline_featured_image":false,"ub_ctt_via":"","footnotes":""},"categories":[1],"tags":[],"class_list":["post-23413","post","type-post","status-publish","format-standard","has-post-thumbnail","category-technology"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 4.9.9 - aioseo.com -->\n\t<meta name=\"description\" content=\"Key Takeaways Context engineeringsupports AI models deliver suitable, reliable results by offering the right information, instructions, memory, and tools for every task. Context engineering vs. prompt engineering comes down to scope: prompt engineering focuses on instructions, while context engineering controls the broader information the model needs for effective working. RAG, fine-tuning, and context window management serve different\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Shaili Gupta\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/\" \/>\n\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 4.9.9\" \/>\n\t\t<meta property=\"og:locale\" content=\"en_US\" \/>\n\t\t<meta property=\"og:site_name\" content=\"Software Development, AI &amp; Technology Blog | Sphinx Solutions -\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"What Is Context Engineering? A Practical Guide for AI Agents\" \/>\n\t\t<meta property=\"og:description\" content=\"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents.\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2026-10-16T06:43:30+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2026-10-16T06:43:30+00:00\" \/>\n\t\t<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n\t\t<meta name=\"twitter:title\" content=\"What Is Context Engineering? A Practical Guide for AI Agents\" \/>\n\t\t<meta name=\"twitter:description\" content=\"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents.\" \/>\n\t\t<script type=\"application\/ld+json\" class=\"aioseo-schema\">\n\t\t\t{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#breadcrumblist\",\"itemListElement\":[{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog#listItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\",\"nextItem\":{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/category\\\/technology\\\/#listItem\",\"name\":\"Technology\"}},{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/category\\\/technology\\\/#listItem\",\"position\":2,\"name\":\"Technology\",\"item\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/category\\\/technology\\\/\",\"nextItem\":{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#listItem\",\"name\":\"Context Engineering: A Detailed Guide\"},\"previousItem\":{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog#listItem\",\"name\":\"Home\"}},{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#listItem\",\"position\":3,\"name\":\"Context Engineering: A Detailed Guide\",\"previousItem\":{\"@type\":\"ListItem\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/category\\\/technology\\\/#listItem\",\"name\":\"Technology\"}}]},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/#organization\",\"name\":\"Software Development, AI & Technology Blog | Sphinx Solutions\",\"url\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/\"},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/author\\\/shaili-gupta\\\/#author\",\"url\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/author\\\/shaili-gupta\\\/\",\"name\":\"Shaili Gupta\",\"image\":{\"@type\":\"ImageObject\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#authorImage\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/d6fe8286f8710717dc303435461506c38eef8d6f74be7af7ce1171dd17db1443?s=96&r=g\",\"width\":96,\"height\":96,\"caption\":\"Shaili Gupta\"}},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#webpage\",\"url\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/\",\"name\":\"What Is Context Engineering? A Practical Guide for AI Agents\",\"description\":\"Key Takeaways Context engineeringsupports AI models deliver suitable, reliable results by offering the right information, instructions, memory, and tools for every task. Context engineering vs. prompt engineering comes down to scope: prompt engineering focuses on instructions, while context engineering controls the broader information the model needs for effective working. RAG, fine-tuning, and context window management serve different\",\"inLanguage\":\"en-US\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/#website\"},\"breadcrumb\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#breadcrumblist\"},\"author\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/author\\\/shaili-gupta\\\/#author\"},\"creator\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/author\\\/shaili-gupta\\\/#author\"},\"image\":{\"@type\":\"ImageObject\",\"url\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/wp-content\\\/uploads\\\/2026\\\/10\\\/Context-Engineering_-A-Detailed-Guide.webp\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#mainImage\",\"width\":750,\"height\":421,\"caption\":\"Context Engineering_ A Detailed Guide\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/context-engineering-guide\\\/#mainImage\"},\"datePublished\":\"2026-10-16T06:43:30+00:00\",\"dateModified\":\"2026-10-16T06:43:30+00:00\"},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/#website\",\"url\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/\",\"name\":\"Software Development, AI & Technology Blog | Sphinx Solutions\",\"inLanguage\":\"en-US\",\"publisher\":{\"@id\":\"https:\\\/\\\/www.sphinx-solution.com\\\/blog\\\/#organization\"}}]}\n\t\t<\/script>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"What Is Context Engineering? A Practical Guide for AI Agents","description":"Key Takeaways Context engineeringsupports AI models deliver suitable, reliable results by offering the right information, instructions, memory, and tools for every task. Context engineering vs. prompt engineering comes down to scope: prompt engineering focuses on instructions, while context engineering controls the broader information the model needs for effective working. RAG, fine-tuning, and context window management serve different","canonical_url":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"BreadcrumbList","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#breadcrumblist","itemListElement":[{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog#listItem","position":1,"name":"Home","item":"https:\/\/www.sphinx-solution.com\/blog","nextItem":{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/#listItem","name":"Technology"}},{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/#listItem","position":2,"name":"Technology","item":"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/","nextItem":{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#listItem","name":"Context Engineering: A Detailed Guide"},"previousItem":{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog#listItem","name":"Home"}},{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#listItem","position":3,"name":"Context Engineering: A Detailed Guide","previousItem":{"@type":"ListItem","@id":"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/#listItem","name":"Technology"}}]},{"@type":"Organization","@id":"https:\/\/www.sphinx-solution.com\/blog\/#organization","name":"Software Development, AI & Technology Blog | Sphinx Solutions","url":"https:\/\/www.sphinx-solution.com\/blog\/"},{"@type":"Person","@id":"https:\/\/www.sphinx-solution.com\/blog\/author\/shaili-gupta\/#author","url":"https:\/\/www.sphinx-solution.com\/blog\/author\/shaili-gupta\/","name":"Shaili Gupta","image":{"@type":"ImageObject","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#authorImage","url":"https:\/\/secure.gravatar.com\/avatar\/d6fe8286f8710717dc303435461506c38eef8d6f74be7af7ce1171dd17db1443?s=96&r=g","width":96,"height":96,"caption":"Shaili Gupta"}},{"@type":"WebPage","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#webpage","url":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/","name":"What Is Context Engineering? A Practical Guide for AI Agents","description":"Key Takeaways Context engineeringsupports AI models deliver suitable, reliable results by offering the right information, instructions, memory, and tools for every task. Context engineering vs. prompt engineering comes down to scope: prompt engineering focuses on instructions, while context engineering controls the broader information the model needs for effective working. RAG, fine-tuning, and context window management serve different","inLanguage":"en-US","isPartOf":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/#website"},"breadcrumb":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#breadcrumblist"},"author":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/author\/shaili-gupta\/#author"},"creator":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/author\/shaili-gupta\/#author"},"image":{"@type":"ImageObject","url":"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Context-Engineering_-A-Detailed-Guide.webp","@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#mainImage","width":750,"height":421,"caption":"Context Engineering_ A Detailed Guide"},"primaryImageOfPage":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/#mainImage"},"datePublished":"2026-10-16T06:43:30+00:00","dateModified":"2026-10-16T06:43:30+00:00"},{"@type":"WebSite","@id":"https:\/\/www.sphinx-solution.com\/blog\/#website","url":"https:\/\/www.sphinx-solution.com\/blog\/","name":"Software Development, AI & Technology Blog | Sphinx Solutions","inLanguage":"en-US","publisher":{"@id":"https:\/\/www.sphinx-solution.com\/blog\/#organization"}}]},"og:locale":"en_US","og:site_name":"Software Development, AI &amp; Technology Blog | Sphinx Solutions -","og:type":"article","og:title":"What Is Context Engineering? A Practical Guide for AI Agents","og:description":"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents.","og:url":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/","article:published_time":"2026-10-16T06:43:30+00:00","article:modified_time":"2026-10-16T06:43:30+00:00","twitter:card":"summary_large_image","twitter:title":"What Is Context Engineering? A Practical Guide for AI Agents","twitter:description":"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents."},"aioseo_meta_data":{"post_id":"23413","title":"What Is Context Engineering? A Practical Guide for AI Agents","description":null,"keywords":null,"keyphrases":{"focus":{"keyphrase":"","score":0,"analysis":{"keyphraseInTitle":{"score":0,"maxScore":9,"error":1}}},"additional":[]},"primary_term":null,"canonical_url":null,"og_title":"What Is Context Engineering? A Practical Guide for AI Agents","og_description":"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents.","og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":"","og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":"What Is Context Engineering? A Practical Guide for AI Agents","twitter_description":"Learn what context engineering is, how it differs from prompt engineering, RAG and fine-tuning, tools and safeguards behind reliable AI agents.","schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"BlogPosting","isEnabled":true},"graphs":[]},"schema_type":"default","schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":"-1","robots_max_videopreview":"-1","robots_max_imagepreview":"large","priority":null,"frequency":"default","local_seo":null,"limit_modified_date":false,"created":"2026-10-15 11:10:22","updated":"2026-10-16 07:48:43","ai":{"faqs":[],"keyPoints":[],"schemas":[],"titles":[],"descriptions":[],"socialPosts":{"email":{"subject":"","preview":"","content":""},"linkedin":[],"twitter":[],"facebook":[],"instagram":[]}},"breadcrumb_settings":null,"seo_analyzer_scan_date":null},"aioseo_breadcrumb":"<div class=\"aioseo-breadcrumbs\"><span class=\"aioseo-breadcrumb\">\n\t\t\t<a href=\"https:\/\/www.sphinx-solution.com\/blog\" title=\"Home\">Home<\/a>\n\t\t<\/span><span class=\"aioseo-breadcrumb-separator\">&raquo;<\/span><span class=\"aioseo-breadcrumb\">\n\t\t\t<a href=\"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/\" title=\"Technology\">Technology<\/a>\n\t\t<\/span><span class=\"aioseo-breadcrumb-separator\">&raquo;<\/span><span class=\"aioseo-breadcrumb\">\n\t\t\tContext Engineering: A Detailed Guide\n\t\t<\/span><\/div>","aioseo_breadcrumb_json":[{"label":"Home","link":"https:\/\/www.sphinx-solution.com\/blog"},{"label":"Technology","link":"https:\/\/www.sphinx-solution.com\/blog\/category\/technology\/"},{"label":"Context Engineering: A Detailed Guide","link":"https:\/\/www.sphinx-solution.com\/blog\/context-engineering-guide\/"}],"featured_image_src":"https:\/\/www.sphinx-solution.com\/blog\/wp-content\/uploads\/2026\/10\/Context-Engineering_-A-Detailed-Guide.webp","author_info":{"display_name":"Shaili Gupta","author_link":"https:\/\/www.sphinx-solution.com\/blog\/author\/shaili-gupta\/"},"_links":{"self":[{"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/posts\/23413","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/users\/21"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/comments?post=23413"}],"version-history":[{"count":5,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/posts\/23413\/revisions"}],"predecessor-version":[{"id":23422,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/posts\/23413\/revisions\/23422"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/media\/23421"}],"wp:attachment":[{"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/media?parent=23413"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/categories?post=23413"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.sphinx-solution.com\/blog\/wp-json\/wp\/v2\/tags?post=23413"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}