Hi everyone,
I’m currently exploring a startup idea and would really appreciate honest feedback from people who work with APIs, documentation, developer tooling, or Postman-based workflows.
The idea is to start from an API spec such as OpenAPI/Swagger, analyze and improve it, then use that improved spec to generate tools for an MCP server dedicated to that API.
On top of that, the system would let technical and non-technical users ask questions about the API through a simple chat interface or endpoint, and get grounded answers based on the spec and related documentation.
The main differentiation I’m exploring is cost efficiency:
-
if multiple users ask the same or semantically similar questions, the system should reuse previous answers instead of making a new LLM call every time
-
if a prompt contains both old and new requests, it should reuse prior context and only call the LLM for the new part
-
the goal is to make API documentation retrieval more scalable and cheaper, while still being accurate and useful
So in short, the concept sits somewhere between:
-
API documentation intelligence
-
MCP generation from API specs
-
AI chat over API documentation
-
semantic caching / partial context reuse for cost optimization
I’d love feedback on a few things:
-
Does this sound like a real problem or more like a nice technical demo?
-
Which part sounds most valuable: improving the API spec, generating MCP tools, or reducing LLM cost for repeated documentation queries?
-
Do you think teams working with APIs would actually adopt something like this?
-
Are there obvious risks or gaps you see right away?
I’d really appreciate any honest thoughts, even critical ones.