Skip to content

Documentation Search

Overview

In this lab, you'll use your AI coding agent to search AWS documentation for information about a recently-launched service. This demonstrates how the MCP Server's knowledge tools give your agent access to up-to-date AWS documentation without you needing to leave your IDE.

What You'll Learn

  • How to prompt your agent to search AWS documentation
  • How the MCP Server retrieves real-time documentation content
  • How to refine documentation searches for more specific results

Instructions

Goal: Ask your agent to search AWS documentation for a recently-launched AWS service and retrieve useful technical details about it.

Info

Hint 1: Try asking about a service launched in the last 12 months — for example, Amazon Bedrock Agents, Amazon Q Developer, or AWS App Studio. The MCP Server has access to current documentation.

Info

Hint 2: Start broad ("What is Amazon Q Developer?") then refine with follow-up questions ("What are the pricing tiers?" or "How does it integrate with IAM?").

Info

Hint 3: Ask your agent to compare two related services or explain when to use one over the other — this tests the agent's ability to synthesize information from multiple documentation pages.

Walkthrough

Click to see the step-by-step walkthrough

  1. Open your AI coding agent in your IDE or terminal.

  2. Ask your agent to search for documentation about a recent AWS service:

"Search the AWS documentation for Amazon Q Developer and summarize its key features."

  1. Review the response — your agent should return information retrieved from official AWS docs via the MCP Server.

  2. Ask a follow-up question to test deeper retrieval:

"What IAM permissions are required to use Amazon Q Developer?"

  1. Try a comparison query:

"Search the AWS documentation and compare Amazon Q Developer with Amazon CodeWhisperer. What are the key differences?"

  1. Verify that the responses contain specific, current information (version numbers, recent feature additions) that wouldn't be available from the agent's training data alone.

Validation

Your progress is tracked automatically. Once your agent has successfully made documentation search calls via the MCP Server, the CloudWatch dashboard will show a ✅ for Module 1.

To verify manually: check that CloudTrail shows recent events with a user agent containing "mcp" — this confirms your agent is communicating through the MCP Server.

Agent-Specific Tips

Claude Code connects to the MCP Server automatically once configured. Documentation searches use the search_documentation tool under the hood. You can ask Claude to explain which tools it used by saying "What MCP tools did you just use?"

Kiro's MCP integration surfaces documentation results inline. After getting a response, you can ask Kiro to dig deeper into specific sections. Kiro will automatically use the read_documentation tool for full page content when summaries aren't sufficient.

Cursor routes MCP requests through its agent mode. Make sure you're in Agent mode (not Ask mode) for the MCP Server tools to be available. If you get generic responses without AWS-specific details, verify your MCP connection is active.

Codex connects to the MCP Server via its tool-use capabilities. Ensure your session has active AWS credentials configured. If documentation searches return empty results, verify the MCP endpoint URL is correctly set in your configuration.