--- title: Why MCP servers should tell agents what to do next description: Many MCP servers act like thin APIs that just hand back data, says AWS principal developer advocate James Ward. He argues a tool response should also tell the... canonical: https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Why MCP servers should tell agents what to do next | daily.dev og:description: Many MCP servers act like thin APIs that just hand back data, says AWS principal developer advocate James Ward. He argues a tool response should also tell the... og:url: https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec og:image: https://api.daily.dev/og/posts/R9zFyO8ec.png og:image:alt: Why MCP servers should tell agents what to do next og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Why MCP servers should tell agents what to do next **[The New Stack](https://daily.dev/sources/newstack)** · 3 min read · 0 upvotes · 0 comments ## Summary Many MCP servers act like thin APIs that just hand back data, says AWS principal developer advocate James Ward. He argues a tool response should also tell the agent what it just did and what to try next, acting like another voice in the conversation. Since MCP has no built-in trace ID, AWS adds a session ID parameter so calls can be stitched into one journey, plus a parameter carrying the user's goal for comparison. AWS opened its MCP Server to everyone in May and added Lambda diagnostics in September. Ward also notes AWS ships a CLI and skills alongside MCP because MCP's OAuth model can't switch AWS user profiles like the CLI can. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How can I trace an agent's full journey across multiple MCP tool calls since MCP has no built-in trace ID? Add a custom session ID parameter to your MCP tools and instruct the agent to pass it on every call, letting you stitch individual calls into one journey. A second parameter can carry the user's original prompt or goal summary, so you can compare intent against the actual path taken and build evals around it. _Anyone wiring up MCP observability can compare approaches and workarounds like this on daily.dev._ ### Should an MCP tool response just return raw data or also suggest what the agent should do next? A tool response should carry guidance alongside the data, telling the agent what it just did, what it might want to do next, and where to find more. AWS principal developer advocate James Ward describes this as treating the MCP server like another voice in the conversation rather than a plain database-style API. _Teams designing agent-facing tool responses can track this evolving MCP design guidance on daily.dev._ ### Why would a team use a CLI or skills instead of MCP for AWS agent tooling? AWS ships a CLI and skills alongside its MCP server partly because MCP's OAuth model cannot toggle between AWS user profiles the way the CLI can. Which approach works best depends on the model being used, the tool descriptions, and how the agent discovers tools, so AWS tests the options against each other with evals. _Developers weighing MCP against CLI-based agent tooling can follow these tradeoffs on daily.dev._ ## Similar posts on daily.dev - [The AWS MCP Server is now generally available](https://daily.dev/posts/the-aws-mcp-server-is-now-generally-available-wonxezacy) · AWS · 1 upvotes · 0 comments - [AWS MCP Server adds a serverless capability for AWS Lambda functions](https://daily.dev/posts/aws-mcp-server-adds-a-serverless-capability-for-aws-lambda-functions-w3yfhlcrx) · AWS · 0 upvotes · 0 comments - [AWS MCP Server Reaches GA with Full API Coverage and IAM-Based Governance](https://daily.dev/posts/aws-mcp-server-reaches-ga-with-full-api-coverage-and-iam-based-governance-gd3gwr6nd) · InfoQ · 1 upvotes · 0 comments - [MCP Servers: Your Next User Is An AI Agent](https://daily.dev/posts/mcp-servers-your-next-user-is-an-ai-agent-apyzw7sof) · marmelab · 2 upvotes · 0 comments - [MCP in action: What makes AI agents actually useful](https://daily.dev/posts/mcp-in-action-what-makes-ai-agents-actually-useful-xzn8qnwi5) · All Things Open · 0 upvotes · 0 comments --- Tags: [#aws](https://daily.dev/tags/aws), [#ai-agents](https://daily.dev/tags/ai-agents), [#observability](https://daily.dev/tags/observability), [#mcp](https://daily.dev/tags/mcp) [View this post on daily.dev](https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Why MCP servers should tell agents what to do next","url":"https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec"},"datePublished":"2026-10-29T16:02:10.464Z","dateModified":"2026-10-29T16:04:58.768Z","description":"Many MCP servers act like thin APIs that just hand back data, says AWS principal developer advocate James Ward. He argues a tool response should also tell the...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/4a60b3fe2546868f0afb9fa6750ec1df?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/4a60b3fe2546868f0afb9fa6750ec1df?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"The New Stack","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"The New Stack","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/newstack","url":"https://daily.dev/sources/newstack"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"aws,ai-agents,observability,mcp","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"The New Stack","item":"https://daily.dev/sources/newstack"},{"@type":"ListItem","position":3,"name":"Why MCP servers should tell agents what to do next"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/why-mcp-servers-should-tell-agents-what-to-do-next-r9zfyo8ec#faq","mainEntity":[{"@type":"Question","name":"How can I trace an agent's full journey across multiple MCP tool calls since MCP has no built-in trace ID?","acceptedAnswer":{"@type":"Answer","text":"Add a custom session ID parameter to your MCP tools and instruct the agent to pass it on every call, letting you stitch individual calls into one journey. A second parameter can carry the user's original prompt or goal summary, so you can compare intent against the actual path taken and build evals around it. Anyone wiring up MCP observability can compare approaches and workarounds like this on daily.dev."}},{"@type":"Question","name":"Should an MCP tool response just return raw data or also suggest what the agent should do next?","acceptedAnswer":{"@type":"Answer","text":"A tool response should carry guidance alongside the data, telling the agent what it just did, what it might want to do next, and where to find more. AWS principal developer advocate James Ward describes this as treating the MCP server like another voice in the conversation rather than a plain database-style API. Teams designing agent-facing tool responses can track this evolving MCP design guidance on daily.dev."}},{"@type":"Question","name":"Why would a team use a CLI or skills instead of MCP for AWS agent tooling?","acceptedAnswer":{"@type":"Answer","text":"AWS ships a CLI and skills alongside its MCP server partly because MCP's OAuth model cannot toggle between AWS user profiles the way the CLI can. Which approach works best depends on the model being used, the tool descriptions, and how the agent discovers tools, so AWS tests the options against each other with evals. Developers weighing MCP against CLI-based agent tooling can follow these tradeoffs on daily.dev."}}]} ```