Integrating Claude Code with WeChat’s New ClawBot: A Step‑by‑Step Guide

Why Connect Claude Code to WeChat

Claude Code is a powerful command‑line tool, but its usage is confined to a terminal. By leveraging WeChat’s newly opened ClawBot API, developers can talk to Claude from any mobile device without installing extra apps, turning the bot into an always‑available AI coding assistant.

Project Overview

The open‑source project qufei1993/cc-weixin (Code Channel for WeChat) implements this integration. The repository is hosted on GitHub and the article walks through its design and implementation.

Architecture: Three‑Layer Decoupling

Layer 1 – WeChat Communication Layer : Handles all interactions with the WeChat server, including HTTP API calls, login credential management, and media upload/download. Upper layers do not need to know any WeChat protocol details.

Layer 2 – Message Polling Layer : The core module monitor.ts uses long polling to fetch new messages, parses them, applies access control, and emits them via the onMessage callback. This layer is platform‑agnostic, so the same code works for other IM services.

Layer 3 – Platform Adaptation Layer : Implemented in server.ts, this MCP server converts WeChat callbacks into a format Claude Code can understand, communicating through stdio without exposing any network ports.

The layered design means that adding support for a different AI tool (e.g., Codex) only requires a new adaptation layer, leaving the first two layers untouched.

Message Handling: Full Media Support

Text messages : Basic Q&A interaction.

Image messages : Claude’s multimodal capability processes screenshots directly.

Voice messages : Automatically transcribed to text before handling.

File transfer : Code or log files can be sent for analysis.

All media files are encrypted with AES; the encryption key is randomly generated for each upload, adding an extra protection layer.

Security Design: Pair Code + Whitelist

To prevent prompt‑injection attacks, cc‑weixin uses a three‑step mechanism:

New users scan a QR code and enter a 6‑digit pairing code (valid for 10 minutes).

Successful pairing adds the user to a whitelist.

The system can be locked in whitelist mode so only authorized users may send messages.

Credential files are protected with chmod 0600, and the pairing code is persisted to disk for cross‑process sharing. The MCP server runs via stdio, exposing no network ports.

Installation and Usage

Prerequisites: Bun runtime and Claude Code with channel support.

# Add the plugin marketplace
/plugin marketplace add qufei1993/cc-weixin

# Install the plugin
/plugin install weixin@cc-weixin

After installation:

Restart Claude Code.

Run /weixin:configure to scan a QR code and bind a WeChat account.

Restart Claude Code with the flag --dangerously-load-development-channels to enable channel listening; the terminal will show “Listening for channel messages from: server:weixin”.

After scanning, the bot sends a 6‑digit pairing code; execute /weixin:access pair <code> to authorize.

Optionally lock the access policy with /weixin:access to enforce whitelist mode.

The whole process takes less than a minute.

Actual Experience

Conversation in WeChat is smooth; response latency is dominated by Claude’s own processing time, while WeChat transmission delay is negligible. Typical use cases include:

On‑the‑go code review by sending a screenshot.

Team members sharing a single Claude Code instance via WeChat.

Voice queries when typing is inconvenient.

Screenshots from both the WeChat side and Claude Code terminal confirm the integration works as described.

Conclusion

WeChat’s release of ClawBot signals its embrace of the AI ecosystem, offering developers a compliant way to build AI applications inside the messaging platform. cc‑weixin demonstrates that connecting AI coding tools to everyday communication tools is straightforward and can fundamentally change developers’ workflows.