What is Mcps

MCPs (Minimal Control Point servers) are a collection of servers designed to wrap public protocol documentation in a semantic-search API, enabling real-time querying of the documentation.

Use cases

Use cases for MCPs include querying protocol documentation in real-time during development, integrating with AI agents for enhanced documentation retrieval, and providing developers with easy access to protocol specifications.

How to use

To use MCPs, navigate to the desired protocol folder and run the corresponding Python script (e.g., python x402_MCP/x402_mcp.py). This will launch the server, allowing you to query the documentation through the API.

Key features

Key features of MCPs include: integration with the cursor/fastmcp tooling stack, pre-built vector stores for efficient search, and the ability to run servers out-of-the-box with included documentation.

Where to use

MCPs can be used in various fields such as software development, API documentation management, and AI applications where semantic search capabilities are beneficial.

MCP Servers

This folder contains a collection of Minimal Control Point (MCP) servers that wrap public protocol documentation in a semantic-search API. They are meant to be run from within the cursor / fastmcp tooling stack so that the rest of this repository (or your own AI agent) can query the docs at runtime.

Directory layout

Each sub-folder contains:

• <name>_mcp.py – the actual server (run it with python <name>_mcp.py)
• <name>_vectorstore.parquet – the embedded vector store generated from the docs
• <name>_docs_full.txt – a plain-text concatenation of the full docs used to create the store

The vector stores (*_vectorstore.parquet) and *_docs_full.txt files are already checked-in for convenience, so you can run the servers out-of-the-box. If you need to regenerate the stores you can run your own offline ingestion pipeline or use the original build utilities (not included in this repository).

# Example: launch the x402 MCP server
python MCPs/x402_MCP/x402_mcp.py  # transport defaults to stdio

Security & secrets

No private API keys, credentials, or wallets are committed to this repository. You may notice that some of the *_docs_full.txt files (and therefore their vector stores) include example or dev-net private keys that come straight from the public documentation of the respective projects (for example, Nillion’s local-devnet keys). Those keys are intentionally public and cannot be used on main-net or test-net systems – they are safe to keep in version control.

If you intend to add your own tooling on top of these MCP servers *do NOT* commit real secrets. Store anything sensitive in an .env file that is listed in the project-level .gitignore.

Requirements

• Python ³⋅¹¹+
• The dependencies pinned in requirements.txt

pip install -r requirements.txt

Contributing

Feel free to open a PR to add additional MCP servers or improve the existing ones. Please verify that any new documentation you ingest is either public-domain or licensed for redistribution and does not contain sensitive credentials.