Public APIs MCP

The Public APIs MCP addresses a common pain point for developers building AI‑powered assistants: quickly discovering and integrating free, publicly available APIs without hunting through documentation or manually parsing lists. By exposing a semantic search interface over the well‑known GitHub collection, this server lets an assistant query for relevant services—such as weather, finance, or translation APIs—and then retrieve full details in a single call. This streamlines the workflow from idea to implementation, allowing assistants to suggest concrete API endpoints and usage patterns in real time.

At its core, the server offers two tools. The first, , performs an embedding‑based search across API names and descriptions. A user can provide a natural language query (“weather data”), and the tool returns ranked matches with concise snippets, each identified by a unique ID. The second tool, , fetches the complete metadata for a chosen API by ID. Together, these capabilities let an assistant first surface options and then drill down into the exact request format, authentication requirements, and example payloads.

Key features include a lightweight, self‑contained index that can be built on demand, meaning the first search will automatically generate embeddings for all catalog entries. The server also exposes two resources: a collection endpoint () and individual API endpoints (), enabling clients to browse or reference APIs directly through the MCP resource protocol. This duality supports both programmatic access (via tools) and exploratory browsing within an assistant’s UI.

Real‑world scenarios abound. A developer might ask their AI helper for “a free API to get current air quality” and receive a ranked list, then request the details of the top match to copy an example curl command. In larger pipelines, a data‑science assistant could automatically pull multiple weather APIs into a data ingestion workflow by iterating over search results. Because the server is open source and MIT‑licensed, teams can host it privately or extend it with custom data sources, ensuring compliance with internal security policies.

Integration is straightforward for any MCP‑compliant client. Once the server is running, a tool call such as can be invoked from within a conversational context; the assistant can then present the results, ask follow‑up questions, and retrieve full details—all without leaving the chat. This tight coupling between semantic search and resource retrieval makes it a powerful addition to any AI assistant that needs to surface external services on demand.