Claude-to-IM-skill

skill
SUMMARY

Bridge Claude Code / Codex to IM platforms — chat with AI coding agents from Telegram, Discord, or Feishu/Lark.

README.md

Claude-to-IM Skill

Bridge Claude Code / Codex to IM platforms — chat with AI coding agents from Telegram, Discord, Feishu/Lark, QQ, or WeChat.

中文文档

Want a desktop GUI instead? Check out CodePilot — a full-featured desktop app with visual chat interface, session management, file tree preview, permission controls, and more. This skill was extracted from CodePilot's IM bridge module for users who prefer a lightweight, CLI-only setup.


How It Works

This skill runs a background daemon that connects your IM bots to Claude Code or Codex sessions. Messages from IM are forwarded to the AI coding agent, and responses (including tool use, permission requests, streaming previews) are sent back to your chat.

You (Telegram/Discord/Feishu/QQ/WeChat)
  ↕ Bot API
Background Daemon (Node.js)
  ↕ Claude Agent SDK or Codex SDK (configurable via CTI_RUNTIME)
Claude Code / Codex → reads/writes your codebase

Features

  • Five IM platforms — Telegram, Discord, Feishu/Lark, QQ, WeChat — enable any combination
  • Interactive setup — guided wizard collects tokens with step-by-step instructions
  • Permission control — tool calls require explicit approval via inline buttons (Telegram/Discord) or text /perm commands / quick 1/2/3 replies (Feishu/QQ/WeChat)
  • Streaming preview — see Claude's response as it types (Telegram & Discord)
  • Session persistence — conversations survive daemon restarts
  • Secret protection — tokens stored with chmod 600, auto-redacted in all logs
  • Zero code required — install the skill and run /claude-to-im setup, or tell Codex claude-to-im setup

Prerequisites

  • Node.js >= 20
  • Claude Code CLI (for CTI_RUNTIME=claude or auto) — installed and authenticated (claude command available)
  • Codex CLI (for CTI_RUNTIME=codex or auto) — npm install -g @openai/codex. Auth: run codex auth login, or set OPENAI_API_KEY (optional, for API mode)

Installation

Choose the section that matches the AI agent product you actually use.

Claude Code

Recommended: npx skills

npx skills add op7418/Claude-to-IM-skill

After installation, tell Claude Code:

/claude-to-im setup

If you want WeChat specifically, you can also say:

帮我接微信

Alternative: clone directly into Claude Code skills

git clone https://github.com/op7418/Claude-to-IM-skill.git ~/.claude/skills/claude-to-im

Claude Code discovers it automatically.

Alternative: symlink for development

git clone https://github.com/op7418/Claude-to-IM-skill.git ~/code/Claude-to-IM-skill
mkdir -p ~/.claude/skills
ln -s ~/code/Claude-to-IM-skill ~/.claude/skills/claude-to-im

Codex

Recommended: use the Codex install script

git clone https://github.com/op7418/Claude-to-IM-skill.git ~/code/Claude-to-IM-skill
bash ~/code/Claude-to-IM-skill/scripts/install-codex.sh

For local development with a live checkout:

bash ~/code/Claude-to-IM-skill/scripts/install-codex.sh --link

The install script places the skill under ~/.codex/skills/claude-to-im, installs dependencies, and builds the daemon.

After installation, tell Codex:

claude-to-im setup

If you want WeChat specifically, you can also say:

帮我接微信桥接

Alternative: clone directly into Codex skills

git clone https://github.com/op7418/Claude-to-IM-skill.git ~/.codex/skills/claude-to-im
cd ~/.codex/skills/claude-to-im
npm install
npm run build

Verify installation

Claude Code: Start a new session and type / — you should see claude-to-im in the skill list. Or ask Claude: "What skills are available?"

Codex: Start a new session and say claude-to-im setup, start bridge, or 帮我接微信桥接.

Updating the Skill

Choose the update flow that matches both your AI agent product and your installation method.

Claude Code

If you installed with npx skills, re-run:

npx skills add op7418/Claude-to-IM-skill

If you installed via git clone or symlink:

cd ~/.claude/skills/claude-to-im
git pull
npm install
npm run build

Then tell Claude Code:

/claude-to-im doctor
/claude-to-im start

Codex

If you installed with the Codex install script in copy mode:

rm -rf ~/.codex/skills/claude-to-im
bash ~/code/Claude-to-IM-skill/scripts/install-codex.sh

If you installed with --link or cloned directly into the Codex skills directory:

cd ~/.codex/skills/claude-to-im
git pull
npm install
npm run build

Then tell Codex:

claude-to-im doctor
start bridge

Quick Start

1. Setup

Claude Code

/claude-to-im setup

Codex

claude-to-im setup

The wizard will guide you through:

  1. Choose channels — pick Telegram, Discord, Feishu, QQ, WeChat, or any combination
  2. Enter credentials — the wizard explains exactly where to get each token, which settings to enable, and what permissions to grant
  3. Set defaults — working directory, model, and mode
  4. Validate — tokens are verified against platform APIs immediately

2. Start

Claude Code

/claude-to-im start

Codex

start bridge

The daemon starts in the background. You can close the terminal — it keeps running.

3. Chat

Open your IM app and send a message to your bot. Claude Code / Codex will respond through the bridge.

When Claude needs to use a tool (edit a file, run a command), you'll see a permission prompt with Allow / Deny buttons right in the chat (Telegram/Discord), or a text /perm command prompt / quick 1/2/3 replies (Feishu/QQ/WeChat).

Commands

All commands are run inside Claude Code or Codex:

Claude Code Codex (natural language) Description
/claude-to-im setup "claude-to-im setup" / "配置" Interactive setup wizard
/claude-to-im start "start bridge" / "启动桥接" Start the bridge daemon
/claude-to-im stop "stop bridge" / "停止桥接" Stop the bridge daemon
/claude-to-im status "bridge status" / "状态" Show daemon status
/claude-to-im logs "查看日志" Show last 50 log lines
/claude-to-im logs 200 "logs 200" Show last 200 log lines
/claude-to-im reconfigure "reconfigure" / "修改配置" Update config interactively
/claude-to-im doctor "doctor" / "诊断" Diagnose issues

Platform Setup Guides

The setup wizard provides inline guidance for every step. Here's a summary:

Telegram

  1. Message @BotFather on Telegram → /newbot → follow prompts
  2. Copy the bot token (format: 123456789:AABbCc...)
  3. Recommended: /setprivacy → Disable (for group use)
  4. Find your User ID: message @userinfobot

Discord

  1. Go to Discord Developer Portal → New Application
  2. Bot tab → Reset Token → copy it
  3. Enable Message Content Intent under Privileged Gateway Intents
  4. OAuth2 → URL Generator → scope bot → permissions: Send Messages, Read Message History, View Channels → copy invite URL

Feishu / Lark

  1. Go to Feishu Open Platform (or Lark)
  2. Create Custom App → get App ID and App Secret
  3. Batch-add permissions: go to "Permissions & Scopes" → use batch configuration to add all required scopes (the setup wizard provides the exact JSON)
  4. Enable Bot feature under "Add Features"
  5. Events & Callbacks: select "Long Connection" as event dispatch method → add im.message.receive_v1 event
  6. Publish: go to "Version Management & Release" → create version → submit for review → approve in Admin Console
  7. Important: The bot will NOT work until the version is approved and published

QQ

QQ currently supports C2C private chat only. No group/channel support, no inline permission buttons, no streaming preview. Permissions use text /perm ... commands. Image inbound only (no image replies).

  1. Go to QQ Bot OpenClaw
  2. Create a QQ Bot or select an existing one → get App ID and App Secret (only two required fields)
  3. Configure sandbox access and scan QR code with QQ to add the bot
  4. CTI_QQ_ALLOWED_USERS takes user_openid values (not QQ numbers) — can be left empty initially
  5. Set CTI_QQ_IMAGE_ENABLED=false if the underlying provider doesn't support image input

WeChat / Weixin

WeChat currently uses QR login, single-account mode, text-based permissions, and no streaming preview.

  1. Run the local QR helper from your installed skill directory:
    • Claude Code default install: cd ~/.claude/skills/claude-to-im && npm run weixin:login
    • Codex default install: cd ~/.codex/skills/claude-to-im && npm run weixin:login
  2. The helper writes ~/.claude-to-im/runtime/weixin-login.html and tries to open it in your browser automatically
  3. Scan the QR code with WeChat and confirm on your phone
  4. On success, the linked account is stored in ~/.claude-to-im/data/weixin-accounts.json
  5. Running the helper again replaces the previously linked WeChat account

Additional notes:

  • CTI_WEIXIN_MEDIA_ENABLED controls inbound image/file/video downloads only
  • Voice messages only use WeChat's own built-in speech-to-text text
  • If WeChat does not provide voice_item.text, the bridge replies with an error instead of downloading/transcribing raw voice audio
  • Permission approvals use text /perm ... commands or quick 1/2/3 replies

Architecture

~/.claude-to-im/
├── config.env             ← Credentials & settings (chmod 600)
├── data/                  ← Persistent JSON storage
│   ├── sessions.json
│   ├── bindings.json
│   ├── permissions.json
│   └── messages/          ← Per-session message history
├── logs/
│   └── bridge.log         ← Auto-rotated, secrets redacted
└── runtime/
    ├── bridge.pid          ← Daemon PID file
    └── status.json         ← Current status

Key components

Component Role
src/main.ts Daemon entry — assembles DI, starts bridge
src/config.ts Load/save config.env, map to bridge settings
src/store.ts JSON file BridgeStore (30 methods, write-through cache)
src/llm-provider.ts Claude Agent SDK query() → SSE stream
src/codex-provider.ts Codex SDK runStreamed() → SSE stream
src/sse-utils.ts Shared SSE formatting helper
src/permission-gateway.ts Async bridge: SDK canUseTool ↔ IM buttons
src/logger.ts Secret-redacted file logging with rotation
scripts/daemon.sh Process management (start/stop/status/logs)
scripts/doctor.sh Health checks
SKILL.md Claude Code skill definition

Permission flow

1. Claude wants to use a tool (e.g., Edit file)
2. SDK calls canUseTool() → LLMProvider emits permission_request SSE
3. Bridge sends inline buttons to IM chat: [Allow] [Deny]
4. canUseTool() blocks, waiting for user response (5 min timeout)
5. User taps Allow → bridge resolves the pending permission
6. SDK continues tool execution → result streamed back to IM

Troubleshooting

Run diagnostics:

/claude-to-im doctor

This checks: Node.js version, config file existence and permissions, token validity (live API calls), log directory, PID file consistency, and recent errors.

Issue Solution
Bridge won't start Run doctor. Check if Node >= 20. Check logs.
Messages not received Verify token with doctor. Check allowed users config.
Permission timeout User didn't respond within 5 min. Tool call auto-denied.
Stale PID file Run stop then start. daemon.sh auto-cleans stale PIDs.

See references/troubleshooting.md for more details.

Security

  • All credentials stored in ~/.claude-to-im/config.env with chmod 600
  • Tokens are automatically redacted in all log output (pattern-based masking)
  • Allowed user/channel/guild lists restrict who can interact with the bot
  • The daemon is a local process with no inbound network listeners
  • See SECURITY.md for threat model and incident response

Development

npm install        # Install dependencies
npm run dev        # Run in dev mode
npm run typecheck  # Type check
npm test           # Run tests
npm run build      # Build bundle

License

MIT

Yorumlar (0)

Sonuc bulunamadi