cccc  by ChesterRa

AI agents co-driving repositories through collaborative planning and critique

Created 5 months ago
265 stars

Top 96.7% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

ChesterRa/cccc is a local-first, multi-agent collaboration kernel orchestrating AI agents for continuous code delivery. It enables AI peers to co-drive repositories, handling planning, building, and critique autonomously. Control is via an interactive Web UI or team chat integrations (Telegram, Slack, Discord), positioning it as a production-minded orchestrator.

How It Works

The 0.4.x rewrite features a daemon-based architecture with a central ccccd daemon managing multiple agent runtimes within "working groups." Each group uses an append-only ledger as its single source of truth, ensuring durable history and reducing latency. Agents interact via an MCP tool surface for reliable message delivery and system control. A mobile-first Web UI serves as the primary control plane, complemented by IM bridges for chat integration. This approach overcomes v0.3.x limitations like fragmented messages and limited actor scaling.

Quick Start & Requirements

  • Requirements: Python 3.9+, macOS/Linux/Windows (WSL for PTY). Requires at least one supported agent runtime CLI.
  • Installation (RC): Via TestPyPI:
    python -m pip install --index-url https://pypi.org/simple \
      --extra-index-url https://test.pypi.org/simple \
      cccc-pair==0.4.0rc16
    
  • Quick Start: Navigate to repo, cccc attach ., cccc setup --runtime <name>, cccc actor add ..., cccc group start, then cccc daemon start and cccc web. Access UI at http://127.0.0.1:8848/.
  • Docs: docs/vnext/README.md, docs/vnext/ARCHITECTURE.md.

Highlighted Details

  • Unified Ledger: Append-only ledger per working group as the single source of truth.
  • N-Actor Model: Groups support multiple independent agent actors for scalable collaboration.
  • MCP Control Plane: Agents manage CCCC state and actors via a defined tool interface.
  • Web-First Console: Responsive UI enables remote management, mobile-friendly.
  • IM Bridge: Integrates with Telegram, Slack, Discord, Feishu/Lark, DingTalk.
  • PROJECT.md Constitution: Agents reference PROJECT.md at scope root as project guidelines.

Maintenance & Community

The project is in Release Candidate (RC) status (v0.4.0rc16), indicating active development and potential breaking changes. Specific community channels or maintainer details were not provided in the excerpt.

Licensing & Compatibility

Licensed under Apache-2.0, permissive for commercial use and closed-source linking.

Limitations & Caveats

The 0.4.x line is RC, with potential breaking changes. Requires a long-lived local daemon (ccccd). Native Windows PTY requires WSL. Remote Web UI access needs robust security (access gateways, authentication).

Health Check
Last Commit

3 days ago

Responsiveness

Inactive

Pull Requests (30d)
31
Issues (30d)
0
Star History
57 stars in the last 30 days

Explore Similar Projects

Starred by Andrew Ng Andrew Ng(Founder of DeepLearning.AI; Cofounder of Coursera; Professor at Stanford), Thomas Wolf Thomas Wolf(Cofounder of Hugging Face), and
4 more.

ag2 by ag2ai

0.5%
4k
AgentOS for building AI agents and facilitating multi-agent cooperation
Created 1 year ago
Updated 2 days ago
Feedback? Help us improve.