# Why MCP servers should tell agents what to do next

DevFeed: [Why MCP servers should tell agents what to do next](<https://devfeed.tech/articles/why-mcp-servers-should-tell-agents-what-to-do-next-106471.md>)

Original publisher: [Read original article](<https://thenewstack.io/aws-mcp-agent-observability/>)

Author: Matthew Burns

Published: 2026-10-29T15:55:39Z

Content type: news

Language: en

Sources: [The New Stack](<https://devfeed.tech/sources/the-new-stack.md>)

Topics: [MSP MCP](<https://devfeed.tech/topics/msp-mcp.md>), [agent to agent communication](<https://devfeed.tech/topics/agent-to-agent-communication.md>), [google-adk](<https://devfeed.tech/topics/google-adk.md>)

Tags: [agent-integration](<https://devfeed.tech/tags/agent-integration.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-foundation](<https://devfeed.tech/tags/agentic-ai-foundation.md>), [agents](<https://devfeed.tech/tags/agents.md>), [agntcon](<https://devfeed.tech/tags/agntcon.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [api](<https://devfeed.tech/tags/api.md>), [apis](<https://devfeed.tech/tags/apis.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-marketplace](<https://devfeed.tech/tags/aws-marketplace.md>), [batch](<https://devfeed.tech/tags/batch.md>), [cli](<https://devfeed.tech/tags/cli.md>), [developer-advocate](<https://devfeed.tech/tags/developer-advocate.md>), [diagnostics](<https://devfeed.tech/tags/diagnostics.md>), [eval](<https://devfeed.tech/tags/eval.md>), [evals](<https://devfeed.tech/tags/evals.md>), [fetching](<https://devfeed.tech/tags/fetching.md>), [guidance](<https://devfeed.tech/tags/guidance.md>), [id](<https://devfeed.tech/tags/id.md>), [mcp](<https://devfeed.tech/tags/mcp.md>), [mcp-servers](<https://devfeed.tech/tags/mcp-servers.md>), [model-context-protocol](<https://devfeed.tech/tags/model-context-protocol.md>), [observability](<https://devfeed.tech/tags/observability.md>), [podcast](<https://devfeed.tech/tags/podcast.md>), [premium-event-coverage](<https://devfeed.tech/tags/premium-event-coverage.md>), [sponsor-aws-marketplace](<https://devfeed.tech/tags/sponsor-aws-marketplace.md>), [sponsored](<https://devfeed.tech/tags/sponsored.md>), [sponsored-podcast](<https://devfeed.tech/tags/sponsored-podcast.md>), [sponsored-premium-event-coverage](<https://devfeed.tech/tags/sponsored-premium-event-coverage.md>), [video](<https://devfeed.tech/tags/video.md>)

## AI overview

James Ward argues that MCP servers should return guidance alongside data, describing what an agent did and possible next steps, rather than acting only as basic APIs. To trace agent journeys, AWS adds a session ID to tool calls and can also pass the user's prompt or goal summary. Teams can compare goals with call paths, build evals, adjust tool descriptions or parameters, and measure task completion. Journey traces can reveal inefficient patterns, such as fetching customer records one at a time, that a batch tool could address. Ward says AWS also offers a CLI and skills: MCP's OAuth model cannot switch AWS user profiles as the CLI can. Which option works best depends on the model, tool descriptions, and discovery, so AWS tests them with evals.

## Source excerpt

Many MCP servers do one thing. An agent asks for data, and the server hands it back. AWS's James Ward The post Why MCP servers should tell agents what to do next appeared first on The New Stack.