Finds what your change left untested, and the device cases nobody runs.
Early AccessEvery agent is free to connect — no card, no checkout.
We use essential cookies to keep you signed in. With your consent we also use non-essential cookies for analytics. Read our privacy policy
Browse › Software Development › Repo Health Check
An MCP server that gives any public GitHub repository a maintainer-health score from 0 to 100, and shows its working.
Repo Health Check runs on FindAgent and connects over MCP — no download, no local runtime. Add the server URL to any MCP-capable client.
https://mcp.findagent.cloud/agents/repo-health-checkChatGPT (Plus / Pro): Settings → Connectors → add a remote MCP server, then paste the URL and approve it.
You add this server inside your client and approve it once — that in-client step is a security check, so there's no one-click button for it.
Signing in: the first time you connect, your client opens a FindAgent sign-in page — approve once and you're connected. There's no API key or token to copy; access is tied to your FindAgent account.
FindAgent points you at this agent’s code and never runs it. It runs on your own computer, with your own access — so it can reach your local tools.
This agent runs code on your machine.
When you run it, it executes with your user’s access — it can read your files, use your local tools, and make network requests, just like any program you run. FindAgent scanned this listing and a person reviewed it, but that is not a guarantee it is safe to run. Review the source before running it.
The source of this local-run agent is private
You can’t inspect the code before running it on your machine. Only run it if you trust the publisher.
Sign in with the CLI once (findagent login), then run:
npx -y @findagent/mcp run repo-health-checkPaste this into your client config (e.g. claude_desktop_config.json or a Cursor mcp.json), then restart the client.
{
"mcpServers": {
"repo-health-check": {
"command": "npx",
"args": [
"-y",
"@findagent/mcp",
"run",
"repo-health-check"
]
}
}
}The sandbox starts with no network access. Its code can connect only to these hosts.
Finds what your change left untested, and the device cases nobody runs.
Runs the intake interview one question at a time and freezes the brief.
Finds the security defects that pass a code review and a green build.
The design spec before the screen gets built: states, copy, accessibility.
Plans an install ad that only claims what the product actually did.
Reviews a mobile change with severities that still mean something.
Keeps a team's written knowledge findable instead of merely large.