Make your MCP server more efficient, in seconds
FieldFlow turns any REST API into a GraphQL-like MCP server so agents fetch only the fields they need.
Built with ❤️ by Guillaume Gay
Spot the Difference
See how FieldFlow swaps painful API rewrites for instant, filtered tools your agents love.
Without FieldFlow
- ❌Rebuild everything in GraphQL
- ❌LLMs ingest raw, noisy API payloads
- ❌Weeks of setup and hand-tuning
- ❌Tokens burned on useless fields
- ❌Agents lost in sprawling responses
- ❌Rigid, hard-to-integrate APIs
With FieldFlow
- ✅No backend changes required
- ✅Only the fields you actually need
- ✅Ready in under two minutes
- ✅Up to 80% fewer LLM tokens
- ✅Automatic field discovery for agents
- ✅Works with any REST API you already ship
Capabilities
Everything you need to ship field-level API tooling from a single OpenAPI specification.
OpenAPI native
Point FieldFlow at any OpenAPI 3.x JSON or YAML file and we discover endpoints, methods, and schemas automatically.
Typed Pydantic models
Generated request and response models preserve aliases, optional fields, and nested refs for precise validation.
Selective responses
Every route accepts an optional fields list so callers fetch only the payload they need—no over-fetching, no manual serializers.
FastAPI gateway
FieldFlow proxies requests with httpx, formats path and query params, and hands back filtered responses through FastAPI routes.
Spec-first fidelity
Recursive schema resolution handles nested objects, arrays, and references so your proxy stays aligned with the upstream API.
MCP tooling optional
Expose the same filtered endpoints to Model Context Protocol clients like Claude Desktop with the generated MCP layer.