Jitapi

What is Jitapi?

Jitapi is an MCP server that lets Claude interact with any API from its OpenAPI spec, using semantic search and a dependency graph to surface only the needed endpoints.

How to use Jitapi?

Install via pip (pip install jitapi) and add the server configuration to your MCP client with command: "uvx" and args: ["jitapi"]. Then register any OpenAPI spec URL, search endpoints in plain English, and Claude chains the right API calls automatically.

Key features of Jitapi

• Zero-config: no API keys required for local embeddings
• Multi-API orchestration: register and query across multiple APIs
• Automatic dependency graph resolution for endpoint chains
• Semantic search in natural language instead of exact paths
• Pluggable embeddings: local, Voyage AI, OpenAI, or Cohere

Use cases of Jitapi

• Querying any REST API (e.g., Stripe, GitHub) via Claude without loading the full spec
• Building cross-API workflows that automatically resolve endpoint dependencies
• Rapid prototyping of integrations without writing a custom MCP server per API

FAQ from Jitapi

How does Jitapi handle large API specs?

It uses semantic search and a dependency graph to surface only what’s needed, avoiding context token waste and hallucinations.

What runtime dependencies does Jitapi require?

Python with pip; it runs via uvx. Local embeddings work out of the box without external API keys.

Can I use my own embedding model?

Yes. Jitapi supports pluggable embeddings: local (default), Voyage AI, OpenAI, or Cohere.

How are endpoint dependencies resolved?

Jitapi builds a dependency graph that knows which endpoints must be called first (e.g., GET /users before POST /orders) and resolves this automatically.

Does Jitapi support multiple APIs simultaneously?

Yes. You can register multiple OpenAPI specs and ask questions that span them.