Pontifex MCP¶
The security and governance layer for MCP servers, built on the official MCP Python SDK.
The problem¶
AI agents are ready to do real work — answer questions from your data, take actions in your systems. MCP (the Model Context Protocol, the open standard agents like Claude use to call tools) gives you a server that exposes those tools. What it doesn't give you is the control that makes a server safe to point at real data: who is calling, what they're allowed to touch, how often, and a record of what happened.
So teams stall in the same place: the AI works in a demo, but nobody will connect it to the orders database, the customer records, or the internal APIs — because there's no authentication, no access control, and no audit trail.
What Pontifex MCP does¶
Pontifex MCP is the connecting layer between AI agents and the systems your business runs on — it turns your existing APIs, data stores, and internal services into governed tools any AI agent can call. That's how you get an AI initiative out of pilot and into the parts of the business that matter, without handing your data to a third party.
Governed is the heart of it: every tool call gets authentication, per-caller scopes, rate limits, observability, and a full audit trail.
It builds on the official MCP Python SDK and stays on open protocols throughout — so you run it on the infrastructure you already use, pair it with any AI vendor, and can strip it out whenever you like. Your data never leaves your environment.
flowchart TB
agent["AI agent<br/>Claude Desktop, your agents"]
subgraph pontifex ["pontifex-mcp "]
gate["Authenticate · check scope · rate limit"]
tool["Your tool"]
audit["Audit log"]
gate --> tool
tool --> audit
end
agent -->|MCP| gate
tool --> adapters["Data adapters"]
adapters --> systems["Your systems<br/>APIs, databases, services"]
Capabilities¶
-
Secure by default
OAuth 2.1 JWTs and
sk_…API keys. Every tool call is authenticated, against any OIDC provider (Auth0, Entra, Clerk, Keycloak). -
Least-privilege scopes
domain: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.
-
Resilient
Per-caller rate limiting, data-source failover, and circuit breaking.
-
Drop-in connectors
Point a config file at an OpenAPI spec and every allowlisted operation becomes a governed tool — no code. Connectors →
Who it's for¶
Reach for pontifex-mcp when you're exposing internal or proprietary systems — an orders API, a
customer database, an analytics warehouse — to AI agents, and unauthenticated tool access isn't an
option. If you're shipping a single public tool over non-sensitive data, the MCP SDK on its own is
simpler.
-
Build with it
Stand up your first authenticated server in a few minutes.
-
Evaluate it
The security model, architecture, and how it fits your stack.
MIT licensed. Part of Argonauts.