cc-connect-ai-bridge（cc 连接 ai 桥）

CC-Connect AI Bridge

Skill by ara.so — Daily 2026 Skills collection.

CC-Connect bridges locally running AI coding agents (Claude Code, Codex, Cursor Agent, Gemini CLI, Qoder CLI, OpenCode, iFlow CLI) to messaging platforms you already use. You can chat with your local agent from Telegram, Feishu, Slack, Discord, DingTalk, WeChat Work, LINE, or QQ — no public IP required for most platforms.

Installation

Via npm (recommended)

# Stable
npm install -g cc-connect

# Beta (includes personal WeChat / Weixin ilink and other beta features)
npm install -g cc-connect@beta

Via binary

# Linux amd64
curl -L -o cc-connect https://github.com/chenhg5/cc-connect/releases/latest/download/cc-connect-linux-amd64
chmod +x cc-connect
sudo mv cc-connect /usr/local/bin/

# macOS arm64
curl -L -o cc-connect https://github.com/chenhg5/cc-connect/releases/latest/download/cc-connect-darwin-arm64
chmod +x cc-connect
sudo mv cc-connect /usr/local/bin/

Build from source (Go 1.22+)

git clone https://github.com/chenhg5/cc-connect.git
cd cc-connect
make build
# binary appears at ./bin/cc-connect

Self-update

cc-connect update          # stable
cc-connect update --pre    # beta / pre-release

Quick AI-assisted Setup

Send this prompt to Claude Code or any AI coding agent — it will install and configure cc-connect automatically:

Follow https://raw.githubusercontent.com/chenhg5/cc-connect/refs/heads/main/INSTALL.md to install and configure cc-connect.

Configuration

mkdir -p ~/.cc-connect
cp config.example.toml ~/.cc-connect/config.toml

Minimal config.toml — Claude Code + Telegram

[project.my-project]
name        = "my-project"
work_dir    = "/home/user/myproject"
data_dir    = "/home/user/.cc-connect/data"
admin_from  = "123456789"          # Your Telegram user ID

[project.my-project.agent]
type = "claude-code"               # claude-code | codex | cursor | gemini | qoder | opencode | iflow

[project.my-project.platform]
type  = "telegram"
token = "$TELEGRAM_BOT_TOKEN"      # set via env or paste value

Multi-project config

# Project 1: Claude on Telegram
[project.dev]
name     = "dev"
work_dir = "/home/user/project-a"
data_dir = "/home/user/.cc-connect/data/dev"

[project.dev.agent]
type = "claude-code"

[project.dev.platform]
type  = "telegram"
token = "$TELEGRAM_BOT_TOKEN"

# Project 2: Gemini on Feishu
[project.research]
name     = "research"
work_dir = "/home/user/project-b"
data_dir = "/home/user/.cc-connect/data/research"

[project.research.agent]
type = "gemini"

[project.research.platform]
type           = "feishu"
app_id         = "$FEISHU_APP_ID"
app_secret     = "$FEISHU_APP_SECRET"
verification_token = "$FEISHU_VERIFICATION_TOKEN"

All supported platform types

All supported agent types

Platform-specific configuration

Telegram

[project.mybot.platform]
type  = "telegram"
token = "$TELEGRAM_BOT_TOKEN"
# Get token from @BotFather on Telegram
# No public IP required — uses long polling

Feishu / Lark

[project.mybot.platform]
type               = "feishu"
app_id             = "$FEISHU_APP_ID"
app_secret         = "$FEISHU_APP_SECRET"
verification_token = "$FEISHU_VERIFICATION_TOKEN"
# Uses WebSocket — no public IP needed
# Set up at https://open.feishu.cn/

Slack

[project.mybot.platform]
type              = "slack"
bot_token         = "$SLACK_BOT_TOKEN"
app_token         = "$SLACK_APP_TOKEN"
# app_token must start with xapp- (Socket Mode)
# Enable Socket Mode in your Slack app settings

Discord

[project.mybot.platform]
type  = "discord"
token = "$DISCORD_BOT_TOKEN"
# Uses Discord Gateway — no public IP needed

DingTalk

[project.mybot.platform]
type       = "dingtalk"
client_id  = "$DINGTALK_CLIENT_ID"
client_secret = "$DINGTALK_CLIENT_SECRET"
# Uses DingTalk Stream — no public IP needed

WeChat Work (WeCom)

[project.mybot.platform]
type           = "wecom"
corp_id        = "$WECOM_CORP_ID"
agent_id       = "$WECOM_AGENT_ID"
secret         = "$WECOM_SECRET"
connection_type = "websocket"   # websocket (no public IP) or webhook

Personal WeChat (beta only)

# Install beta build first
npm install -g cc-connect@beta

# Scan QR code to log in
cc-connect weixin setup

[project.mybot.platform]
type = "weixin"
# Auth state is stored after QR scan — no public IP needed

Running cc-connect

# Default config location: ~/.cc-connect/config.toml
cc-connect

# Custom config path
cc-connect --config /path/to/config.toml

# Run specific project only
cc-connect --project my-project

Chat Commands (sent in the messaging platform)

Slash command examples (in chat)

# Ask the agent to review a PR
Review the changes in branch feature/auth and summarize the risks.

# Switch to a different model
/model gemini-2.0-flash

# Change working directory
/dir /home/user/other-project

# Schedule a daily summary
/cron 0 6 * * * Summarize GitHub trending Go repos and post here

# Read agent memory
/memory

# Update agent memory
/memory Always prefer table-driven tests in Go. Use context.Context for cancellation.

Multi-Agent Relay (group chat orchestration)

Bind multiple bots in one group chat so they communicate with each other:

# Bot 1: Claude
[project.claude-bot]
name     = "claude-bot"
work_dir = "/home/user/project"
data_dir = "/home/user/.cc-connect/data/claude"

[project.claude-bot.agent]
type = "claude-code"

[project.claude-bot.platform]
type  = "telegram"
token = "$TELEGRAM_CLAUDE_BOT_TOKEN"

# Bot 2: Gemini (same group)
[project.gemini-bot]
name     = "gemini-bot"
work_dir = "/home/user/project"
data_dir = "/home/user/.cc-connect/data/gemini"

[project.gemini-bot.agent]
type = "gemini"

[project.gemini-bot.platform]
type  = "telegram"
token = "$TELEGRAM_GEMINI_BOT_TOKEN"

Add both bots to the same Telegram group. Mention them by name and they will pick up and respond to each other's output.

Voice and Multimodal

For platforms that support voice (Telegram, Feishu, DingTalk), enable speech in config:

[speech]
provider = "openai"           # openai | azure | google
api_key  = "$OPENAI_API_KEY"

[speech.stt]
model = "whisper-1"

[speech.tts]
model = "tts-1"
voice = "alloy"

Send a voice message in Telegram → cc-connect transcribes it → forwards text to the agent → optionally speaks the response back.

For images/screenshots: just attach the image in chat. cc-connect forwards it to multimodal-capable agents.

Cron Scheduling

# In config.toml, per-project cron settings
[project.mybot.cron]
timeout = "30m"          # max runtime per job
fresh_session = true     # start a new session for each run

From chat:

# Every weekday at 9am, run a standup summary
/cron 0 9 * * 1-5 Summarize open PRs and blockers in this repo

# Every hour, check for new issues labeled "urgent"
/cron 0 * * * * Check for new GitHub issues labeled urgent and notify me

Admin configuration

[project.mybot]
admin_from = "alice,123456789"   # comma-separated usernames or IDs

Admins can use /shell, /dir, and other privileged commands. Non-admins cannot.

Directory structure

~/.cc-connect/
├── config.toml                          # main config
└── data/
    └── projects/
        └── my-project.state.json        # persisted dir override, session state

Common patterns

Pattern 1: Mobile-first AI development

Install cc-connect on your dev machine, connect it to Telegram. From your phone:

1. Send a task description as a message
2. Agent runs on your machine
3. Streamed response appears in chat
4. Reply to continue the conversation

Pattern 2: Team shared AI bot (Slack/Feishu)

One cc-connect instance per project, shared in a team channel. Team members can ask questions, trigger code reviews, or run analyses — all without terminal access.

Pattern 3: Scheduled reports

/cron 0 8 * * 1 Generate a weekly summary of commits and open issues, format as markdown

Pattern 4: Multi-project routing

Run multiple projects in one cc-connect process, each bot handles a different repo or concern.

Troubleshooting

Bot not responding

1. Check that cc-connect is running: ps aux | grep cc-connect
2. Verify the platform token/credentials in config
3. For Telegram: ensure the bot token is valid (/start in BotFather chat)
4. For Feishu: check WebSocket connection in Feishu Open Platform console

Agent not found

# Verify agent is installed and in PATH
which claude     # for claude-code
which gemini     # for gemini
which codex      # for codex

Set explicit path in config if needed:

[project.mybot.agent]
type = "claude-code"
bin  = "/usr/local/bin/claude"

Session inherits broken state

Use /continue which forks the session, or start fresh:

/session new

Personal WeChat not available

You need the beta build:

npm install -g cc-connect@beta
cc-connect weixin setup   # scan QR in terminal

Context too long / auto-compress

Enable in config:

[project.mybot.agent]
auto_compress        = true
compress_threshold   = 80000   # tokens before compression kicks in

Config not found

Default path is ~/.cc-connect/config.toml. Pass explicitly:

cc-connect --config /custom/path/config.toml

Check version

cc-connect --version

Environment variables reference

Values in config.toml can reference env vars using $VAR_NAME syntax.

Platform setup guides

• Feishu/Lark: docs/feishu.md
• DingTalk: docs/dingtalk.md
• Telegram: docs/telegram.md
• Slack: docs/slack.md
• Discord: docs/discord.md
• WeChat Work: docs/wecom.md
• Personal WeChat (beta): docs/weixin.md
• QQ / QQ Bot: docs/qq.md