A spec-to-PR build agent where no PR ships unverified.

Learn more → · Install · Roadmap

Why

I built Nightcrew because I wanted an autonomous agent that could build from my backlogs overnight without me having to trust its word that the work is done. Coding agents are happy to open a pull request and declare success, and I found that checking those claims by hand costs almost as much as writing the code myself. Nightcrew hard-blocks every pull request behind a verification gate that re-runs the tests independently, checks that the builder didn't weaken them, and makes a judge from a different model family cite concrete evidence for every acceptance criterion, so the PR that lands in front of me carries its own proof.

Features

• Verification gate — an independent test re-run, a tamper check that blocks removed or weakened tests, and an evidence judge that maps each acceptance criterion to concrete evidence in the diff before any PR can open.
• Judge never shares a family with the builder — the gate refuses to run if the judge and builder resolve to the same model family, so the builder can't grade its own homework.
• Pluggable dispatch seam — builder and judge are config-selected command templates (claude -p, codex exec, any CLI) or OpenRouter calls, with no vendored SDK.
• Spec as the contract — each repo carries a nightcrew.yaml of items with acceptance criteria; underspecified items are skipped and logged rather than guessed at.
• Evidence-table PRs, never auto-merged — the PR body is a criterion-by-criterion evidence table with the run cost, and merging is always left to a human.
• Caps and a build log — every run is recorded to SQLite, and daily or per-run dollar caps stop the run cleanly at the item boundary.

Quick start

git clone https://github.com/SeanL128/nightcrew && cd nightcrew && uv sync
uv run nightcrew enroll /path/to/your/repo
uv run nightcrew run

Requires Python 3.11+, uv, git, an authenticated gh CLI, and at least one dispatch backend (the claude or codex CLI, or an OPENROUTER_API_KEY).

Configuration

Option	What it does	Default
nightcrew.yaml → config.test_cmd	The repo's test command; enrollment runs it live and refuses repos where it fails, and the gate re-runs it on every build	required
nightcrew.yaml → items[]	Spec items with id, description, acceptance_criteria, deps	required
fleet.json → repos[]	Enrolled repos with path, priority, trust, max_items_per_run	required
fleet.json → dispatch.roles	Per-role backend: a command template or an OpenRouter model, with an optional explicit family	{}
fleet.json → caps	daily_usd, per_run_usd, max_items run limits	none

More in docs/USAGE.md.

Roadmap

• Full pipeline: pick → plan → build → gate → fix → PR
• Hybrid gate: independent test re-run, tamper check, cross-family evidence judge
• Fleet CLI: enroll / run / status / digest with a SQLite build log and dollar caps
• Portable scheduling recipes: GitHub Actions, cron, systemd (see examples/)
• First public release (v0.1.0)
• Value learning and a trust ratchet, gated on real outcome data; PRs stay propose-only until then

License

MIT

---

Built by Sean Lindsay · seanlindsay.xyz