AST-verified structure · AI-written descriptions
Docs that track your code.
Point Docloom at your GitHub repo and get API docs built from your actual code — endpoints, request bodies, response shapes, query parameters — hosted on a clean docs site, regenerated whenever your code changes.
GitHub sign-in · read-only access · no credit card
Not ready to sign in? Try it on any public repo — no account
How it works
01
Connect a repo
Sign in with GitHub and pick a repository — public or private. Read-only access, nothing is ever written back to your code.
02
We parse, then we write
A TypeScript AST parser extracts the structural facts — endpoints, request bodies and their fields, response shapes, query parameters, headers — and AI writes descriptions on top. Facts come from your code, never from the model.
03
Docs, hosted
Structured markdown, published to a clean docs site on your own subdomain. Regenerate on merge and preview the diff before it goes live.
What ends up in your docs
Every fact below is extracted from your source by a compiler — before any AI touches the page.
Request bodies
Zod schemas are resolved to plain fields — types, constraints, defaults, optionality — even when the schema lives in another file.
Response shapes
What each endpoint returns: JSON payloads, field types, and the HTTP status as written in your code, per branch.
Query & header inputs
searchParams and header reads with names as written — typed when a schema in your repo describes them.
Honest gaps
Anything the parser cannot prove renders as “not documented in source” — never invented, never a guess dressed up as a fact.
FAQ
Will it hallucinate my API?
Structural facts (endpoints, request-body fields, response shapes, query parameters) come from deterministic AST parsing of your code — the AI only writes natural-language descriptions on top of that structure. And when the parser can't prove something from source, the docs say “not documented in source” instead of inventing it. Treat the prose as a strong first draft worth reviewing, not an infallible oracle.
Do you store my source code?
No. We fetch your repo, process it in memory to generate docs, and discard the raw code. We never retain source beyond the active processing session.
What frameworks does it support?
Next.js App Router route handlers (route.ts / route.tsx). If a repository uses Express, Fastify, or NestJS instead, the docs say so explicitly rather than showing an empty result.