feat: Amazon Bedrock AgentCore MCP Server

[theumbrella1]

Summary

This PR creates a MCP server for Amazon AgentCore. It is inspired by Strands Agents MCP Server, which is based on Langchain mcpdoc. It has two mcp tools, search_docs (returns relevant top k docs) and fetch_doc (fetches the full content from a doc url). We'll have a llm.txt dynamically generated by mkdocs for AgentCore documentation.

Note: default_factory of llm_texts_url in config.py needs to be updated after finalizing llm.txt location

Changes

• MCP server with two tools, search_docs and fetch_doc
• Data classes and util files for the tools
• Docusaurus changes including md file, server-cards.json, and sidebars.ts
• README (generated by Kiro)
• Unit tests (generated by Kiro)
• Docstrings (generated by Kiro)

User experience

User can use MCP inspector to run the server locally. It can also be easily installed on a client of user's choosing, such as Kiro, Q CLI, Cursor, and Claude Code, as a mcp server:

{
  "mcpServers": {
    "awslabs.amazon-bedrock-agentcore-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.amazon-bedrock-agentcore-mcp-server@latest"],
      "env": {
        "FASTMCP_LOG_LEVEL": "ERROR"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Example use cases include:

• “What tools do you have access to?”

• “Convert this code to agentcore”/ “Transform to agentcore” / “ Make this agentcore compatible”

• “Troubleshoot errors”/“do you see any errors on cloudwatch?”

• “What agents can I invoke?” / “can you test agentv3”/ “say hello to strandsv4 agent”

• “Create a dynamodb gateway” (downloads smithy model directly and deploys)

• “what tools are available in this gateway?”

• “Can you find me some good examples for...“ / ”I need an example for hosting strands on runtime“

• “What agents can I invoke?” / “Is strandsagentv6 ready to invoke?”

• deploy to runtime with oauth

• Invoke runtime with oauth / “can you generate a bearer token”

• list memory resources I can use

• create memory resource / use this memory in my agent code

Checklist

If your change doesn't seem to apply, please leave them unchecked.

• I have reviewed the contributing guidelines
• I have performed a self-review of this change
• Changes have been tested
• Changes are documented

Is this a breaking change? N

RFC issue number:

Checklist:

• Migration process documented
• Implement warnings (if it can live side by side)

Acknowledgment

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.

[Vivekbhadauria1]

A major reconsideration is to reuse SQLite etc instead of implementing doc indexer/fetcher/cacher.

[Vivekbhadauria1]

why these?

[Vivekbhadauria1]

add tags for agents and AI agents, agentic AI etc

[Vivekbhadauria1]

why not have llms.txt in the same package to start with? Eventually, we can use https://github.com/aws/bedrock-agentcore-starter-toolkit/tree/main/documentation github page?

[ryanycoleman]

Is that about logistics in hosting the llms.txt file at a later date or do you see an advantage in statically packaging the file? I don't prefer the potential for drift if the customer does not update the MCP Server.

[Vivekbhadauria1]

This can be moved to a singleton class for better testing and readability.

[Vivekbhadauria1]

looks very involved, why not use SQLite instead?