Tools
Everything here either collects the data behind the reports or lets you check it yourself. The open-source tools are MIT licensed. All of them work from public registries, public source code and published documents, and none connects to a running system it does not own.
On this site
Built on the weekly sweep. The data behind both is open at /data.
Search and filter the catalogued MCP servers by what they do, how they are maintained and how they scored in the weekly sweep.
How the public MCP registries compare with what is actually in the source repositories, updated with each sweep.
The Major Labs data served over the Model Context Protocol, so an agent can query the census and the scoreboards directly.
Run it yourself
The instruments, published so the numbers can be checked.
The same static checks behind the security scoreboard, as a one-line GitHub Action for your own MCP server. It reads your source and never connects to anything.
The scanner behind the weekly census. It reads public registries and public source code, and it never connects to a running server.
A verification service for agent payment mandates: present a mandate and it returns a signed allow or deny with a reason code and a receipt anyone can re-verify.
The five kits
Open-source primitives for governing an AI agent, one for each question: who, may, spends, did, remembers.