groundctl

Always know what to build next.

npm install -g @groundctl/cli
★ Star on GitHub npm

You can run 5 AI agents in parallel.
But you can't keep track of what they're doing.
groundctl fixes that.

When you scale agents:

Core loop

$ groundctl claim "auth-system"
  ✓ claimed by this session

$ groundctl claim "api-v2"
  ✓ claimed by this session

# no overlap. no confusion.

$ groundctl next
  → markets-de (high priority, no blocking deps)

$ groundctl complete "auth-system"
  ✓ released — ready for next agent

How it works

①
claim
Agent reserves a work unit. No other agent can take it.
②
complete
Marks it done. Releases the claim. State updates instantly.
③
next
Surfaces what to build next — by priority, with no blocking deps.
groundctl watch runs in background after groundctl init.
After every Claude Code session: ingests transcript → updates shared state → regenerates PROJECT_STATE.md + AGENTS.md.
The next agent reads them and knows exactly where to start. Zero manual steps. Always in sync.

Get started

$ npm install -g @groundctl/cli
$ cd your-project
$ groundctl init

  ✓ Database ready
  ✓ Claude Code hooks installed
  ✓ PROJECT_STATE.md generated
  ✓ groundctl watch started

# Or bootstrap from existing git history:
$ groundctl init --import-from-git
groundctl claim <feature> reserve a work unit for this session
groundctl complete <feature> mark done, release claim
groundctl next what should the agent build next
groundctl status macro view of what's built
groundctl agents list active agents with stale detection
groundctl dashboard visual cockpit — LE PLAN · LE CHANTIER · CORPS DE MÉTIER

Works with your orchestrator

Native
Claude Code Agent Teams
PROJECT_STATE.md and AGENTS.md are loaded automatically by every teammate at startup. No export needed.
Native
Codex CLI
Hooks, watcher, and JSONL parser built-in. groundctl watch detects Codex sessions automatically.
Export
Conductor
Generates a task list you can use as reference when setting up Conductor workspaces.
$ groundctl export --conductor
  → .conductor/tasks.md

Conductor runs your agents.
groundctl tells them what to build.

The problem is not that AI agents are weak.
It's that humans can't track them when they scale.
That's what groundctl solves.

groundctl builds itself

groundctl was built using groundctl.

18 sessions 29 features 100% implemented

PROJECT_STATE.md was updated automatically after every session.
Each agent read it on startup and knew exactly where to start.

View PROJECT_STATE.md →