OPEN API support for non SSE mcp and fast api

The idea is:

Allow users to import an OpenAPI (Swagger) JSON file—especially from FastAPI and MCPO—to automatically generate a custom API integration node/tool within N8N. This node should behave like an agent-ready tool and support:

  • :white_check_mark: Bearer token authentication, ideally managed via environment variables or credential stores
  • :white_check_mark: Multiple endpoints and operations (GET, POST, PUT, DELETE, etc.)
  • :white_check_mark: A configurable base URL to point to local/dev/prod API environments
  • :white_check_mark: Input/output schema mapping (parameters, request bodies, and responses)
  • :white_check_mark: Optional query/path/header parameters mapped dynamically from the spec
  • :white_check_mark: Agent-like tool mode to be used in workflows with LLM or decision logic

This would align N8N with modern agent architectures (e.g., OpenAI tool calling, LangChain agents, etc.) and dramatically improve interoperability with FastAPI/MCPO projects.


My use case:

I’m building APIs using FastAPI and MCPO that expose AI tools, backend services, and microservice endpoints. These APIs already generate a valid openapi.json. I want to:

  • Drop the OpenAPI spec into N8N
  • Have it auto-generate a node for every endpoint (with inputs/outputs mapped)
  • Use it securely with Bearer token authentication
  • Connect it to other tools (e.g., LLM nodes, database automation, webhooks)
  • Use it as part of an agent system that selects the right tool based on context

Right now, this requires lots of manual API node configuration. An OpenAPI parser would eliminate that burden and speed up development dramatically.


I think it would be beneficial to add this because:

  • :brain: Reduces friction for developers integrating with existing APIs
  • :locked: Enables secure, token-based access without leaking credentials
  • :high_voltage: Accelerates automation building by removing repetitive configuration
  • :puzzle_piece: Fits perfectly with N8N’s goal of no-code/low-code extensibility
  • :robot: Opens the door for agent-based and LLM toolchains in N8N
  • :counterclockwise_arrows_button: Follows the same principle used in OpenAI tool calling and RAG pipelines

This is a foundational step to make N8N agent-compatible and usable as a flexible, AI-ready orchestrator.


Any resources to support this?


Are you willing to work on this?

Yes, I’d be happy to contribute or test with FastAPI-based APIs, especially those exposed through MCPO.