Open source API archaeology
Every app already documents its API. It just does it in the network tab.
Drop a HAR file or point OakParse at a site. It groups the endpoints, infers the schemas, finds the auth, follows values from one call into the next, and hands you a playground and an OpenAPI spec.
HAR gives deeper results
A HAR holds real request and response bodies, so schemas, auth, and value flow come out grounded in what actually happened. URL discovery can only see what a site publishes.
Drop one or more .har files
Parsed in your browser. Nothing is uploaded.
Endpoint grouping
Collapses traffic into templated routes like /users/{id} from the shape of the paths.
Value flow
Follows a token from the login response into the header of every call after it.
OpenAPI and Postman
Exports a spec built from observed traffic, with discovered routes marked as such.
Playground
Full request editor with snippets in five languages. Sends only when you press Send.
Flows
Detects login, refresh, search, pagination, checkout, upload, and more.
NVIDIA NIM
Answers questions against your real endpoint list. It cannot invent an endpoint.
HAR parsing, schema inference, and every export run entirely in your browser. Only URL discovery, the playground send, and the assistant reach the server, and credentials are stripped before anything is passed to NVIDIA NIM. The API key stays server side and is never shipped to the browser.