ShipRepo OpenAPI

AI that reads your repo, fixes the bug, runs tests in a sandbox, and opens a pull request.

This repo hosts the public OpenAPI 3.1 spec for the ShipRepo API plus copy-paste recipes in your favourite language.

👉 Try it in 30 seconds — no signup: https://shiprepo.com/demo.html

---

Why ShipRepo

Most "AI coding tools" are a chat window around a model. ShipRepo is an automation layer — a real multi-agent pipeline that runs in a hardened Docker sandbox and delivers a ready-to-merge pull request.

What others do	What ShipRepo does
Generate code in a chat	Plan → Locate → Fix → Review → Test → Open PR
Leave you to copy-paste	Push a branch, open the PR, ping reviewers
Hope the model got it right	Retry loop + debug step + stronger-model escalation
One model, one prompt	Per-step model routing + BYOK + best-of-N on retries

---

Quickstart

1 · Grab a demo token (free, no card)

curl -sS -X POST https://api.shiprepo.com/v1/demo/token \
  -H "Content-Type: application/json" \
  -d '{"email": "you@company.com"}'

You get an sr_demo_... token valid for 2 tasks on public repos — Planner → Locator → Inspector only, so you can eyeball the Locator accuracy without burning credits.

2 · Create a task

curl -X POST https://api.shiprepo.com/v1/tasks \
  -H "Authorization: Bearer $SHIPREPO_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "repo_url": "https://github.com/expressjs/express",
    "prompt":   "find the file handling multipart uploads"
  }'

3 · Stream the logs

curl -N https://api.shiprepo.com/v1/tasks/{id}/logs/stream \
  -H "Authorization: Bearer $SHIPREPO_TOKEN"

Server-Sent Events: log (per step), state (every ~1.5s), end (done).

---

Spec files

• openapi-public.yaml — the public spec (tasks, streaming, BYOK, demo)
• examples/node.js — end-to-end Node.js
• examples/python.py — end-to-end Python + sseclient
• examples/bash.sh — curl + jq
• examples/github-action.yml — issue-label → PR

---

Generate an SDK

The spec is OpenAPI 3.1, works with any generator:

# TypeScript
npx @openapitools/openapi-generator-cli generate -i openapi-public.yaml -g typescript-fetch -o sdk-ts/

# Python
npx @openapitools/openapi-generator-cli generate -i openapi-public.yaml -g python -o sdk-py/

# Go
npx @openapitools/openapi-generator-cli generate -i openapi-public.yaml -g go -o sdk-go/

Or drop it into Postman / Insomnia / Scalar in one click.

---

What's under the hood

Every task runs through:

clone → Planner → Locator (grep + import graph) → Inspector
      → Fixer (best-of-N on Pro/Scale retries) → Reviewer
      → Validator (install · typecheck · lint · security · test)
      → debug loop (up to 2 retries, escalate to stronger model)
      → Git commit · push · PR (if GitHub connected)
      → save repo memory + accepted diff for next run

Sandbox is locked down: non-root user, read-only rootfs + tmpfs, dropped capabilities, network=none, 512 MB memory, 3-minute hard timeout, auto-destroyed.

Full details at https://shiprepo.com/docs.html.

---

Contributing

Found a typo or want to suggest an endpoint? PRs welcome.

The spec itself is maintained server-side and regenerated on each release — cosmetic PRs here get merged quickly.

---

Links

• Dashboard — https://cp.shiprepo.com
• Docs — https://shiprepo.com/docs.html
• Full API reference — https://shiprepo.com/api.html
• Interactive Swagger — https://api.shiprepo.com/docs
• Demo (30 sec, no card) — https://shiprepo.com/demo.html
• Changelog — https://shiprepo.com/changelog.html

---

© ShipRepo · MIT (the spec — the platform is commercial)