Link to your material

https://github.com/serverless-dna/powertools-mcp

Description

This project implements an MCP server that enables Large Language Models (LLMs) to search through AWS Lambda Powertools documentation.

The server accesses the live documentation search_index.json data and re-constructs a local search index using lunr.js. This provides an identical search experience for AI Agents and returns the exact same results as a person would get on the website. With the index being local searches are super fast and the index is cached for the life of the server to save rebuilding used indexes. Since the MCP Server uses real search data it is capable of working for any Powertools document site so naturally supports ALL the runtimes.

It provides 2 tools for AI:

search_docs: For searching a mkdocs index (first time use will create the session cached index)

fetch_doc_page: A simple webscraper that fetches a Powertools document URL page and returns the main content div as markdown for the AI to consume. Only pages from Powertools document sites can be loaded by this tool (all other domains are rejected).

I hope you find this useful!

Preferred contact

(Optional) Social Network

No response

(Optional) Additional notes

I have released this as the "Unofficial Powertools MCP Server" because its not endorsed by AWS and I want to respect that the AWS Powertools team should control the projects.

I am more than happy to donate this to the Powertools team as an official project starting point if there is interest.

Other important note about this: I am researching and plan to contact mkdocs to see if they are interested in me creating a core project that baselines the mkdocs MCP Server and officially leverage their index code from mkdocs (right now it is reverse engineered because the search_index.json is custom for the way mkdocs sets up its search engine confiuration via the plugin.

Acknowledgment

• I understand this content may be removed from Powertools for AWS Lambda (Python) documentation if it doesn't conform with the Code of Conduct