A community-driven registry for Claude, Cursor, Windsurf, Cline & more. Not affiliated with Anthropic.
Are you the author? Sign in to claim
AI agent that implements your specs, opens PRs, and reviews code overnight. Queue GitHub issues or PRDs, wake up to pull
Overnight PRD execution for AI-native devs and small teams

Night Watch is an async execution layer for well-scoped engineering work. It takes PRDs or queued board items, runs Claude CLI or Codex in isolated git worktrees, and opens pull requests while you're offline.
Think of it as a repo night shift: you define the work during the day, Night Watch executes the queue overnight, and you wake up to PRs, review fixes, QA output, and audit results.
Night Watch is built for AI-native solo developers, maintainers, and small teams that already work from specs. It is not trying to replace every engineering workflow or turn software delivery into a hands-off black box.
# 1. Install globally
npm install -g @jonit-dev/night-watch-cli
# 2. Initialize in your project
cd your-project
night-watch init
# 3. Check provider detection
night-watch run --dry-run
# 4. Add a well-scoped ticket (board mode) or PRD (file mode)
night-watch board create-prd "Implement feature X" --priority P1
# OR: echo "# My First PRD\n\nImplement feature X..." > docs/prds/my-feature.md
# 5. Run or install cron
night-watch run # Run once
night-watch install # Setup automated cron
New to Night Watch? Follow our 5-Minute Walkthrough to go from zero to your first AI-generated PR.
Queue work as GitHub issues and track them through Draft → Ready → In Progress → Review → Done.

See all scheduled agent runs across your projects at a glance.

# Quick start
npm install -g @jonit-dev/night-watch-cli
cd your-project
night-watch init
night-watch doctor
echo "# My First PRD\n\n## Problem\n..." > docs/prds/01-my-feature.md
night-watch run
External agents and automation should use the JSON-first manageability commands instead of scraping human output:
night-watch agent status --json
night-watch health --json
night-watch config list --json
night-watch config get reviewerEnabled --json
night-watch config set reviewerEnabled false --json
night-watch job pause reviewer --json
night-watch job resume reviewer --json
JSON modes write one JSON document to stdout on success and reserve stderr for errors. night-watch status --json keeps its legacy shape for backward compatibility; night-watch agent status --json adds a stable wrapper with schemaVersion, status, paused, queue, board, health, and lastRuns.
Paused jobs are stored in night-watch.config.json under pausedJobs. Cron scripts and queue claim/dispatch paths read this state before starting work, so night-watch job pause <job> --json prevents new cron/queue-dispatched runs for that job until resumed.
Night Watch is strongest when:
Night Watch is a weaker fit when:
| Provider | CLI Command | Auto-Mode Flag | Slash Commands |
|---|---|---|---|
claude | claude | --dangerously-skip-permissions | -p "/command-name" |
codex | codex | --yolo | --prompt "text" |
claude--provider codex flag or "provider": "codex" in configNight Watch supports passing custom environment variables to the provider CLI via the providerEnv config field. This lets you point the Claude CLI at any Anthropic-compatible endpoint — including GLM-5.
Add providerEnv to your night-watch.config.json:
{
"provider": "claude",
"providerEnv": {
"ANTHROPIC_API_KEY": "your-glm5-api-key",
"ANTHROPIC_BASE_URL": "https://your-glm5-endpoint.example.com"
}
}
These variables are:
night-watch run, night-watch review)night-watch install, so automated runs also pick them up--dry-run output for easy debugging| Use Case | Environment Variables |
|---|---|
| GLM-5 via custom endpoint | ANTHROPIC_API_KEY, ANTHROPIC_BASE_URL |
| Proxy / VPN routing | HTTPS_PROXY, HTTP_PROXY |
| Custom model selection | Any provider-specific env var |
See Configuration > Provider Environment for full details.
npm install -g @jonit-dev/night-watch-cli
npx @jonit-dev/night-watch-cli init
git clone https://github.com/jonit-dev/night-watch-cli.git
cd night-watch-cli
yarn install && yarn build && npm link
Full documentation is available at nightwatchcli.com.
| Document | Description |
|---|---|
| 5-Minute Walkthrough | From zero to your first AI-generated PR |
| Troubleshooting | Common errors and how to fix them |
| Commands Reference | All CLI commands and their options |
| Privacy | Anonymous telemetry and opt-out details |
| Document | Description |
|---|---|
| Ticket/PRD Format | Board tickets and PRD files, dependencies, lifecycle |
| Configuration | Config file, env vars, CLI flags, providerEnv, notifications, schedules |
| Web UI | Dashboard pages and interface reference |
| Document | Description |
|---|---|
| Architecture Overview | System design, execution flows, data persistence |
| CLI Package | Command structure and integration |
| Core Package | Domain logic, DI container, repositories, config loader |
| Server & API | REST endpoints, SSE, middleware, multi-project mode |
| Scheduler Architecture | Global job queue, dispatch modes, cross-project balancing |
| Build Pipeline | TypeScript → esbuild bundling, CI/CD, publishing |
| Bash Scripts | Cron execution scripts and worktree management |
| Document | Description |
|---|---|
| Dev Onboarding | Getting started as a contributor |
| Local Testing | Test CLI locally without publishing |
| Contributing | Development workflow, conventions, PR process |
Night Watch collects anonymous product telemetry by default to understand usage and improve the CLI. Disable it anytime with night-watch telemetry disable, NW_TELEMETRY_DISABLED=1, or DO_NOT_TRACK=1. See Privacy for the exact collected and never-collected lists.
| Document | Description |
|---|---|
| AI-Driven Org Analysis | How Night Watch maps to AI-driven org principles |
MIT License - see LICENSE for details.
npx CLI installing 100+ agents, commands, hooks, and integrations in one command
干净、强大、属于你的 AI Agent 平台 --AI agents, without the clutter.
Native macOS app to monitor Claude AI usage limits and watch your coding sessions live
Pocket Flow: Codebase to Tutorial