v2.0.0 — Open-source launch

Not affiliated with Anthropic. A community-built command center for Claude Code.

CCC v2 is built around a single bet: don't own execution, attach to it. Every Claude Code session running on your Mac shows up on the kanban, regardless of how it was launched.

Why v2

v1 was an opinionated wrapper. v2 inverts the relationship: Claude Code's own on-disk state is the source of truth. The dashboard is a view onto it. You can run zero, one, or 50 sessions however you want — terminal, headless, dashboard-spawned — and they all show up.

What's in v2

Kanban with GitHub Issues

Open issues from your repo come in as cards. Start session spawns a headless Claude attached to the issue. Cards move through:
ISSUES → WORKING → REVIEW → IN TESTING

Worktree view

Every branch your sessions are touching, with PR badges, commit/push state, file sizes, and time-gap markers across days.

Sibling-session commit coordination

Multiple terminals on the same clone? CCC ships a scratch chat file the sessions read/write to negotiate who commits next — so five sessions on main don't clobber each other.

Per-turn auto-summaries

Every turn produces a DID / INSIGHT / NEXT STEP USER block, so you can scan 20 active sessions in 2 minutes without reading transcripts.

Hand-off to terminal or Claude Desktop

Any session card has Launch in terminal and Open in Claude Desktop buttons. Sessions live headless — attach and detach freely.

Spawn CCC-orchestrated sessions from a terminal

A new session inherits the previous one's context automatically when you spawn it via the orchestration mode.

Install

git clone https://github.com/amirfish1/claude-command-center
cd claude-command-center
./run.sh

Open http://localhost:8090./ macOS, Python 3 stdlib, no external deps beyond gh (optional, for GitHub integration).

Demo

▶ [2:46 walkthrough on YouTube](https://youtu.be/_WRf0hH6yhg)

Compatibility

- macOS only for now (Linux PRs welcome — file an issue with what breaks)
- Tested with Claude Code's terminal CLI and headless mode
- Optional integrations: gh (GitHub), vercel (deploy status)

What's next

Top of the backlog (file an issue if you want to nudge priority):
- Linux support
- Multi-machine / shared state across remote dev boxes
- Direct Anthropic API mode (no Claude Code dependency)
- More hand-off targets (Cursor, Zed, JetBrains)

Thanks

Issues, PRs, ideas welcome. Especially interested in edge cases that break the on-disk-state assumption.