ApiLot

Inspiration

API documentation gets outdated fast, especially during hackathons where teams ship quickly and change endpoints constantly. We wanted to remove the manual burden of keeping Postman collections in sync and make API docs update themselves as code evolves.

What it does

Apilot is an autonomous API documentation agent for backend projects. It scans codebases, detects API route changes, classifies breaking vs non-breaking updates, syncs docs to Postman, and stores change history in Neo4j.
It supports realtime mode (watch) and one-shot mode (sync) for teams that don’t want to keep a watcher running.

How we built it

We built Apilot as a Go CLI with Cobra.
Core flow:

Extract routes from source code (Gin, Express, FastAPI/Django patterns + AI fallback/provider switch).
Diff routes against saved state to detect added/modified/deleted endpoints.
Classify breaking change impact using OpenAI.
Sync endpoint docs/examples to Postman collections.
Persist changeset timeline in Neo4j for auditability and project history.

We added quality-of-life features like:

Auto workspace/collection setup in Postman
Re-init safety prompts and cleanup
Retry logic for Postman rate limits/5xx
DEBUG mode for full AI/process traces
Table-formatted history output

Challenges we ran into

Postman API quirks (strict schema expectations and occasional 400/429/500 behavior)
Balancing AI power with API cost/performance
Avoiding unnecessary sync calls for non-API file edits
Making extraction robust across different frameworks and project structures
Keeping terminal UX clean while still giving actionable debug info

Accomplishments that we're proud of

End-to-end autonomous API doc sync from code changes to Postman
Multi-framework parsing with AI-assisted enrichment
Breaking-change intelligence + graph-based history in Neo4j
Strong resilience: retries, pending sync queue, and offline reconciliation
Clean CLI experience (init, watch, sync, diff, history)

What we learned

Reliable developer tools need deterministic logic first, AI second
API integrations need defensive engineering (backoff, retries, graceful fallbacks)
Good DX is as important as core functionality (clear logs, progress, safe re-init)
Hybrid architecture (parsers + AI + stateful sync) is powerful for real-world automation