DEV Community

matengtian
matengtian

Posted on

Bridge Feishu/Lark Chat to Claude Code or Codex CLI for Real-Time AI Coding

#feishu #lark #claude #codex

Imagine debugging a production issue while your team is pinging you on Feishu or Lark. You switch to your terminal, run Claude Code or Codex CLI, fix the bug, then paste the solution back into chat. That back-and-forth kills flow. Enter Lark Coding Agent Bridge – a bot that connects your Feishu/Lark messenger directly to local Claude Code or Codex CLI sessions.

What Problem Does It Solve?

Developers who use AI coding agents (like Claude Code or Codex CLI) often work in isolated terminals. When a teammate asks for help or a code change in chat, you have to context-switch: read the request, open a terminal, run the agent, wait for output, then copy-paste results back. This bridge eliminates that friction. It creates per-chat sessions, so each conversation gets its own agent context. It also supports multiple workspaces, meaning you can have separate agent instances for different projects.

How to Use It

  1. Install the bridge on your machine (see the tool’s docs for setup).
  2. Invite the bot to a Feishu/Lark group or chat.
  3. Send a command like /code fix the login bug in auth.js.
  4. The bot streams the agent’s response as a rich card in real time – no more waiting for a wall of text.
  5. Each chat keeps its own session, so you can have multiple ongoing conversations with different contexts.

Example interaction:

User: /code refactor the database query to use JOINs
Bot: (streaming card) Analyzing query...
Bot: (streaming card) Here's the optimized SQL:
SELECT u.name, o.total FROM users u JOIN orders o ON u.id = o.user_id;
Bot: (streaming card) Done. Want me to test it?
Enter fullscreen mode Exit fullscreen mode

Why It’s Interesting

  • Streaming cards – no more waiting for full responses; you see progress as the agent thinks.
  • Per-chat sessions – context is preserved per conversation, so you can have separate threads for different tasks.
  • Multiple workspaces – run agents for different codebases simultaneously.
  • Open-source bridge – you control the infrastructure; no data leaves your machine.

This tool is perfect for teams that use Feishu/Lark as their daily driver and want to integrate AI coding assistants without leaving their chat app. It turns your messenger into a collaborative coding terminal.

👉 Get Lark Coding Agent Bridge

Top comments (0)