The security, access-control, and governance layer for MCP servers.
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"pontifex-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# pontifex-mcp The security and governance layer for MCP servers, built on the official [MCP Python SDK](https://github.com/modelcontextprotocol/python-sdk). `pontifex-mcp` lets you build [MCP](https://modelcontextprotocol.io) servers that connect AI agents to real systems without giving up control over who can call what. You write the tools; it handles authentication, per-caller scopes, rate limits, and a full audit trail. ## Key features - **Secure by default** — OAuth 2.1 JWTs *and* `sk_…` API keys; every tool call is authenticated. Any OIDC provider (Auth0, Entra, Clerk, Keycloak). - **Least-privilege scopes** — `namespace:resource:action`, checked before every call. Callers can't widen their own access. - **Auditable** — every call recorded: who, what, when, data source, cache hit, latency. - **Standards-based** — RFC 9728 discovery + `WWW-Authenticate`; MCP clients bootstrap auth on their own. - **Resilient** — per-caller rate limiting, adapter failover, circuit breaking. - **Observable** — Logfire / OpenTelemetry tracing and metrics wired in. - **Drop-in connectors** — generate governed tools from an OpenAPI spec (code or config), with optional per-user OAuth token exchange (RFC 8693) to the downstream. - **Built on the MCP SDK** — keep its tools, protocol, and transports; add the controls a production server needs. - **Coding-agent friendly** — bundles an official agent skill (`uvx library-skills`) so your coding agent builds on guidance that matches your installed version. Asymmetric-only JWT validation, generic auth errors, and no token claim can escalate a caller. ## Install ```bash pip install pontifex-mcp # or: uv add pontifex-mcp ``` Requires Python 3.12+. The floor below needs nothing else; Postgres and Redis come in only when you turn on API-key auth. ## Start in a few lines `PontifexMCP` is a drop-in subclass of the MCP SDK's `FastMCP`. The floor needs no database, no Redis, and no auth — the caller is anonymous and every ca…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.