MCP Proxy sits between your AI agents and the tools they call, reviewing every request in real time and connecting safely to any MCP server you run.
How it works
The resolver authenticates every call: OIDC, a static token, GitHub, or AWS STS.
Client and server traffic run the same ordered check chain; the first non-allow verdict wins.
A blocked call gets a JSON-RPC error; the session survives unless integrity is violated.
The check chain, in order
{
"method": "tools/call",
"params": { "name": "db_delete_all" }
}{
"error": {
"code": -32010,
"message": "policy denied"
}
}Why it’s safe to run
A blocked action returns a clear error and work keeps moving.
Restricted tools never show up, so there’s nothing risky to pick.
Guardrails, masking, audits, and identity plug into your setup.
Need it built into your own platform?
MCP Proxy is a library, not just a standalone app. Plug in your own tools and identity provider, and run it anywhere your agents do.