cc-connect

// Overview

cc-connect is a Go daemon that sits between a coding agent already installed on your machine and a chat platform you already use. The agent keeps running locally, with your repositories, your credentials and your shell; the chat app becomes its interface. That means you can inspect or modify code, troubleshoot a failure, run commands and kick off automation from your phone, without exposing the machine to the internet — most supported platforms connect outbound over WebSocket, long polling or socket mode, so no public IP is required.

Agent coverage is broad: Claude Code, Codex, Cursor Agent, Gemini CLI, Qoder CLI, OpenCode (Crush), iFlow CLI, Kimi CLI, Pi and GitHub Copilot are supported directly, plus anything that speaks the Agent Client Protocol (ACP) — which is how Devin is supported. Goose and Aider are listed as planned. Platform coverage runs to Feishu/Lark, WPS Xiezuo, DingTalk, Slack, Telegram, Discord, WeChat Work, Weibo, LINE, QQ (NapCat/OneBot), the official QQ Bot, Matrix and personal WeChat via ilink; LINE is the one that needs a public webhook URL.

Control happens through slash commands in the chat: /model to switch models, /reasoning to tune reasoning effort, /mode to change permission mode, /memory to read and write the agent's instruction files, and /dir <path> to change where the next session starts (with /dir <number> and /dir - for history jumps). Privileged commands such as /dir and /shell are gated on an admin_from list of user IDs configured per project; /whoami and /status report the ID you are calling from. A /dir reset restores the configured work_dir and clears the persisted override.

Beyond one agent in one chat, it handles multi-bot relay — bind several bots in a group chat and let them talk to each other, so you can ask Claude something and pull in Gemini on the same thread. One process serves multiple projects, each with its own agent and platform pairing. Scheduled tasks are written in natural language ("every day at 6am, summarize GitHub trending"). Voice messages and screenshots are handled with STT/TTS and multimodal forwarding, and the interface ships in English, Simplified and Traditional Chinese, Japanese and Spanish.

// What it does

• Bridges 10+ local coding agents — Claude Code, Codex, Cursor Agent, Gemini CLI, Kimi CLI, OpenCode, Qoder, iFlow, Pi, Copilot — plus any ACP-compatible agent
• 13 chat platforms including Slack, Telegram, Discord, Matrix, Feishu, DingTalk and WeChat Work; most need no public IP
• Slash-command control from chat: /model, /reasoning, /mode, /memory, /dir, /status, /whoami
• Multi-bot relay — several agents in one group chat, able to answer each other
• Multi-project: one process, several projects, each with its own agent and platform pairing
• Natural-language scheduled tasks
• Voice and image input with STT/TTS and multimodal forwarding
• Web admin UI for visual configuration, and a five-language interface

// Getting started

Order matters: cc-connect bridges an agent CLI, so install and authenticate at least one agent first. Starting cc-connect without one exits with an error such as "claudecode: claude CLI not found in PATH" and the web UI never comes up.

brew install --cask claude-code            # macOS / Linux Homebrew
# or
npm install -g @anthropic-ai/claude-code   # any platform

npm install -g @openai/codex               # OpenAI Codex

npm install -g cc-connect

# or
brew install cc-connect

# or build from source
git clone https://github.com/chenhg5/cc-connect.git
cd cc-connect
make build

cc-connect web

# or manual configuration
mkdir -p ~/.cc-connect
cp config.example.toml ~/.cc-connect/config.toml
vim ~/.cc-connect/config.toml

./cc-connect

cc-connect update         # stable
cc-connect update --pre   # include pre-releases

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Checking on a long-running coding agent from your phone while away from the desk
• Letting a team drive a shared coding agent from a Slack or Feishu channel
• Running scheduled agent jobs described in plain language, with output delivered to chat
• Comparing two agents' answers side by side in one group conversation via multi-bot relay

// How cc-connect compares

cc-connect alongside other open-source coding-agent harnesses & runtimes tools AI/TLDR tracks, ranked by GitHub stars.