Does this write the documentation for me?

It builds the prompt, not the final docs. Paste the prompt plus your actual code or OpenAPI spec into an LLM and the model produces the reference content. This keeps your source private and lets you use any model you prefer.

Why specify the authentication method?

Auth shapes the request examples and the headers every endpoint needs. Telling the model whether you use Bearer tokens, API keys, OAuth, or basic auth ensures the generated curl and code samples actually work.

Can it produce OpenAPI or Swagger output?

Yes. The prompt asks the model to align with OpenAPI 3.1 conventions, including path parameters, request bodies, and typed response schemas, so the output drops cleanly into a spec or a docs site.

What should I paste alongside the prompt?

Paste your route handlers, controller code, or an existing partial spec. The more concrete the source, the fewer assumptions the model makes about field names, types, and status codes.

Will it document error responses?

Yes. The generated prompt explicitly requests an error table with status codes, error codes, and remediation guidance so consumers know how to handle 4xx and 5xx responses.

API Documentation Prompt Builder

API documentation prompt builder

Good API docs are mostly mechanical: every endpoint needs a description, a parameter table, an auth note, a working request example, and a list of possible responses. This builder turns your endpoint inventory into a single structured prompt that tells an LLM exactly how to produce OpenAPI-aligned reference documentation with real code samples — so you spend minutes reviewing instead of hours writing.

How it works

You list each endpoint (method plus path), describe the parameter types, choose an authentication method, and add a couple of example use cases. The tool assembles a prompt that instructs the model to document each endpoint with a summary, a typed parameter table, request and response examples in curl and one language of your choice, and an error table. Paste the prompt together with your actual source code or spec into any LLM. Everything is generated in your browser; your code is never uploaded to this tool.

Tips and examples

Paste real source. Give the model your controller or route code so it uses your exact field names and status codes rather than inventing them.
Pick the right auth. Bearer, API key, OAuth, and basic auth each change the request examples — choosing correctly avoids broken curl snippets.
Add edge-case use cases. Listing a pagination call or a validation failure prompts the model to document the non-happy paths developers actually hit.
Iterate per endpoint. For large APIs, run the prompt in batches of five to ten endpoints to keep output detailed and within context limits.