← From the lab

API integration Beta

API MCP

Make an existing API usable as tools in an AI client, starting from its OpenAPI description.

From an API operation to a tool.

API MCP reads an OpenAPI 3.x document and exposes selected operations through the Model Context Protocol (MCP). Each exposed operation becomes a tool that forwards a request to the underlying HTTP API.

It runs locally and is useful for connecting an API you already use to an MCP-compatible client. Choose the API description, configure its destination and credentials, then decide which operations the client should be able to call.

Start with a limited set of operations.

Managed configurations start read-only. Named features or individual endpoints can be enabled explicitly, with separate policies for different account profiles. Credentials can come from environment variables or, on supported macOS builds, Keychain.

The macOS approval flow asks locally before policy changes and non-read requests. These controls do not establish that an API description is accurate or that an operation is appropriate for your data. Inspect the contract and test with non-production credentials before allowing writes.

API MCP is a public beta and an independent project, without affiliation to the API providers it can connect. Platform support and setup details belong in its repository. Billy MCP is a concrete integration built around the same approach.

Ask me about API MCP →