CCC
English · 日本語 · 简体中文
Your coding agents outgrew your terminal.
CCC is a free, local command center for coding agents. Every Claude Code, Codex,
Cursor, Grok, Devin and Kimi session lands on one live board. Each row tells you
what the agent did and what it needs from you, so you answer the one that's
blocked and leave the rest working. Not affiliated with Anthropic.
Stop babysitting terminals
- See who needs you. Working, waiting, done: every row says what the agent did and what it's asking for. You often don't have to open it.
- Answer in one tap. When a session stops with a question, you get the options with a summary of each. Pick one and it keeps going.
- Plan with the smart model, build with cheap ones. An Opus planner can hand slices to Codex, Gemini (via Antigravity), Grok or Devin and review every diff.
- Never lose a session. Close the dashboard and your agents keep running. Search everything they ever said.
- Reply from your phone, in the same live view.
curl -fsSL https://raw.githubusercontent.com/amirfish1/claude-command-center/main/scripts/install.sh | CCC_FROM=readme bash
Needs Git and Python 3.9+ (macOS or Linux; Windows). Opens http://localhost:8090 and asks before installing a background service. Sends one anonymous daily ping; CCC_TELEMETRY_DISABLED=1 turns it off (what it sends).
Or brew install amirfish1/tap/ccc && ccc, or grab the macOS DMG.
Want to look first? The live demo runs in your browser with fake data, no install.
CCC v5.41 on demo data. The refunds session split its tests out to Codex and its review to a second Claude; both lanes report back under it.
“Hey Amir, great product. I tried about 20 before finding yours. I have been really enjoying it.”
CCC user
Works with the agents you already use
Nine engines on one board, however you launched them: from CCC, a terminal, or
an IDE. CCC reads each agent's own session files, so there's nothing to wrap
or migrate. Local and source-available: no account, no cloud, free to use at
work (FSL-1.1-MIT).
When an agent stops to ask, you see it
When you hit your usage limit, the fleet keeps going
Start free in 5 minutes
New to coding agents? The first-run wizard checks your machine, helps install
Node, the Claude Code CLI, and a local free-model router, then runs your first
real task on a $0 model. You approve each install step. It ends with the
honest math: "This run cost $0. At API prices it would have cost $X."
Walk through setup · Free providers and privacy.
Free providers have their own limits. The keyless provider logs prompts for
training; CCC asks for consent before enabling it. Do not send it private code.
See CCC at work
One board, nine engines
See sessions you started in a terminal or from CCC. Each agent's own files
keep the list up to date.
Flow canvas & Project tree
Group sessions by the work they belong to. Keep your live sessions next to
a map of your projects.
Split conversations
Drag a session onto the edge of an open conversation to read two agents
side by side. Each has its own input bar.
Find anything, from any session
Search your session history without extra setup. An optional semantic mode
helps when you cannot remember the exact words.
Sessions that coordinate without you
Post once to a group chat and every participant gets the message. No copying
answers between terminals.
Durable queues, one inbox
Put work in a WatchTower queue.
Tickets survive closed sessions, and workers can pick them up in parallel.
Workers that specialize over time
Workers read their queue's shared learnings before a task and write back
afterwards, so the next worker has that experience.
Work from anywhere
Monitor sessions and reply from your phone on a trusted network. CCC stays
local by default. Settings > Phone access guides you through Tailscale.
Phone setup.
All captures use seeded demo data. Full feature guide.
Quickstart
You need Git and Python 3.9+ for the dashboard. To launch agents, use an
installed agent CLI or let the first-run wizard help set one up. gh is
optional for GitHub features.
Prefer Python tools? Build and run a wheel with uvx or pipx.
This is a source-build preview, not a PyPI release.
WatchTower comes with it
WatchTower powers CCC's queues. See setup and requirements.
Running on Windows
irm https://raw.githubusercontent.com/amirfish1/claude-command-center/main/scripts/install.ps1 | iex
From a clone, run .\run.ps1. Native Windows runs in the foreground; WSL2
supports the Linux systemd service route. Windows guide.
From source
git clone https://github.com/amirfish1/claude-command-center
cd claude-command-center
./run.sh
Open http://localhost:8090, then pick a repo before
starting work. Keep the terminal open for a foreground launch.
Running on Linux
Use ./run.sh in the foreground or ./run.sh --install-service for a systemd
user service. Linux and WSL2 guide ·
macOS services · Docker.
Your agent config
CCC asks before registering hooks or installing skills in your agent config.
You can skip them, review the diff, or remove them later in Settings.
What changes, and how to undo it.
Engine support
All nine engines can start sessions from CCC. Follow-up and model controls
vary; Kilo Code has no resume, and Cursor IDE sync is metadata-only.
GitHub Copilot CLI and VS Code Copilot Chat are read-only,
as are Devin cloud sessions. Full engine support matrix.
Why this exists
CCC attaches to the work you already have instead of requiring you to launch
every session through it. Close the dashboard and your agents keep running.
How attachment works.
How it compares
CCC is a session dashboard and coordination tool. Routers provide the model
backends it can use. Comparisons and scope.
Share your fleet
A real share card from one of the maintainer's machines. Yours comes from the throughput page: totals only, no prompts, code, or project names.
Recent
v5.41: Routine / Deep model picks that weigh cost and your remaining usage, helper lanes as real rows under their parent, model prices in the picker, voice switching mid-read, and ccc restart.
See CHANGELOG.md and releases.
Choose Watch > Releases on GitHub for update notifications.
Docs
Community
Join the Discord server (#general-fleet, #bugs-and-ideas, #ship-log) or browse GitHub Discussions.
Contributing
See CONTRIBUTING.md. Feedback and bug reports are welcome.
License
Functional Source License 1.1, MIT Future License (FSL-1.1-MIT) © 2026
Amir Fish. Free to use, modify, and run at work, including on your own servers
for your team. You may not sell CCC or offer it as a competing product or hosted
service. Each release becomes MIT two years after it ships. Versions released
before 2026-07-28 remain under the MIT License; some third-party
contributions stay MIT (see NOTICE).
Acknowledgments
Built on Claude Code.
The gh and Vercel CLIs provide optional integrations.
