← all repositories
ChesterRa/cccc

Your coding agents need a group chat with read receipts

Most multi-agent coding setups stay fragile demos because coordination lives in terminal scrollback with no delivery guarantees; CCCC replaces that with durable messaging and read receipts for agents.

1.1k stars Python AgentsCoding Assistants
cccc
Velocity · 7d
+2.1
★ / day
Trend
cooling
star history

What it does

CCCC is a local-first Python daemon that turns disjointed coding agents into a coordinated group with durable state. It replaces best-effort terminal prompting with an append-only ledger.jsonl that records every message and event, plus explicit read cursors, acknowledgments, and reply-required tracking so you know exactly which agent saw what. A web dashboard, CLI, MCP tools, and IM bridges all talk to the same single-writer daemon, while runtime data stays in CCCC_HOME and out of your repository.

The interesting bit

The clever angle is not the chat interface but the messaging semantics: recipient routing via @foreman and @peers, attention acknowledgments, and auto-wake for disabled agents. It treats agent coordination as a distributed systems problem—complete with delivery obligations and a single-writer log—rather than a UI convenience.

Key highlights

  • Orchestrates 12 first-class runtimes (Claude Code, Codex CLI, Gemini CLI, ChatGPT Web, Grok Build, and others) in one group, each with distinct roles.
  • All state lives in an append-only JSONL ledger; the daemon is the sole writer, eliminating race conditions.
  • Runs entirely offline after a local install with zero external dependencies—no database, broker, or containers required.
  • Remote supervision via IM bridges (Telegram, Slack, Discord, Feishu, DingTalk, WeCom, Weixin) and a local web UI.
  • ChatGPT Web can join as a real actor through browser delivery and remote MCP, though it needs a public HTTPS tunnel.

Caveats

  • ChatGPT Web integration requires exposing the daemon through a public HTTPS URL; ChatGPT rejects localhost, plain HTTP, and private tailnets, and GPT-5.x Pro sessions currently cannot function as local-development runtimes at all.

Verdict

Reach for CCCC if you are trying to run multiple coding agents as a durable 24/7 workflow rather than a fragile demo. If you are happy with one terminal and one agent, this is overkill; the value is in the coordination, not the chat window.

Frequently asked

What is ChesterRa/cccc?
Most multi-agent coding setups stay fragile demos because coordination lives in terminal scrollback with no delivery guarantees; CCCC replaces that with durable messaging and read receipts for agents.
Is cccc open source?
Yes — ChesterRa/cccc is open source, released under the Apache-2.0 license.
What language is cccc written in?
ChesterRa/cccc is primarily written in Python.
How popular is cccc?
ChesterRa/cccc has 1.1k stars on GitHub and is currently cooling off.
Where can I find cccc?
ChesterRa/cccc is on GitHub at https://github.com/ChesterRa/cccc.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.