HPNC — Human-Planned Night Crew

Early Development

This project is under active development. APIs, CLI interfaces, and configuration formats may change at any time. Not yet recommended for production use.

Humans plan by day. Agents implement by night.

HPNC builds the orchestration and quality assurance layer that turns AI coding from interactive babysitting into autonomous overnight execution.

How it works

1. Plan — structure work as stories with machine-readable frontmatter using BMad
2. Queue — add stories to the night queue: hpnc queue add story.md
3. Validate — pre-flight checks: hpnc validate
4. Run — start the night run: hpnc start
5. Review — check results in the morning: hpnc status

Quick install

pip install hpnc
hpnc init

Requirements: Python 3.12+, Git 2.20+, Claude Code and/or Codex CLI

Architecture

Story File (.md)
  → Queue Manager (parse frontmatter)
  → Validator (check night-ready)
  → Dispatcher (orchestrate)
  → Task Runner:
      → Git (create worktree + branch)
      → AgentExecutor (implement)
      → AgentExecutor (review)
      → Quality Gates (build, tests, lint)
      → Event Listener (write run.yaml)
  → Report Generator
  → CLI (morning report)

For AI Agents

Need a single context file for your AI agent? Download HPNC.md — auto-generated from these docs, raw markdown, LLM-optimized.