hachiwii

hachiwii / twinny

Public

A Bridge between Feishu/Lark and CodeX

25
1
80% credibility
Found May 28, 2026 at 25 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Twinny is a bridge that connects an AI coding assistant (Codex) to Feishu/Lark chat. It runs on the owner's local computer and lets them and their team chat with the AI directly in Feishu groups. The owner installs Twinny, sets up a Feishu bot, and then anyone in configured groups can have conversations with the AI assistant, create topics for different tasks, and collaborate on coding work together. Security is built in with options for private or team-wide access, and sensitive information is automatically protected.

How It Works

1
💬 Someone hears about Twinny

A developer or team member learns about Twinny through word of mouth, GitHub, or a friend who uses it to connect their AI assistant to their team's chat.

2
🤖 They want to chat with their AI assistant in Feishu

Instead of opening separate tools, they want to talk to their AI coding assistant directly through their team's Feishu group chat, just like messaging a colleague.

3
🛠️ They install Twinny on their computer

They run a simple installer that sets everything up on their Mac or Linux machine, connecting their Feishu bot to their AI assistant.

4
📱 They create a Feishu bot for their team

Following the setup guide, they create a bot app in Feishu and connect it to Twinny, so messages from their team can reach the AI assistant.

5
✨ The magic moment: chatting with AI in Feishu

Now their team can send messages to the bot in their Feishu group, and the AI assistant responds, creates code, reviews work, and explains things—all within the familiar chat interface.

6
Different ways teams use it
🔒
Private mode

Only the owner can chat with the AI, keeping work private and secure

🌐
Team mode

Anyone in the group can chat with the AI, enabling collaborative coding sessions

🎉 The team has an AI coding buddy in their chat

Now the whole team can ask questions, get code reviews, and work on projects together—all through Feishu messages, with the AI assistant running safely on the owner's computer.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 25 to 25 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is twinny?

Twinny connects Feishu/Lark, the Chinese enterprise messaging platform, to Codex, Anthropic's CLI-based AI coding assistant. You run it as a local daemon on macOS or Linux, configure a Lark bot with the right permissions, and then chat with that bot to drive Codex tasks. Instead of opening a terminal, you send commands like `/thread fix the login bug` or `/goal implement user authentication` from a Lark group, and Codex works in your local workspace. The daemon handles message routing, Codex thread management, workspace isolation per conversation, and sends results back to Lark as formatted messages with status cards.

Why is it gaining traction?

The appeal is obvious if you live in Lark for team communication: you get Codex-powered development without switching contexts. The command interface is surprisingly rich for a bridge tool. You can fork conversation threads, run background goals, inject messages into active Codex turns with `/steer`, and watch documents for @bot mentions that route into active threads. Multiple isolated profiles let you run guest or team Codex sessions without touching your personal `~/.codex`. The setup is guided by an interactive installer, and secrets stay in your Keychain or `auth.json` rather than the config file.

Who should use this?

Teams already deep in Feishu/Lark for daily standups and async communication who also want AI-assisted coding through Codex. If your workflow involves discussing features in Lark threads and then manually copy-pasting prompts into a terminal, Twinny eliminates that friction. It's less useful if your team uses Slack, Discord, or Teams, or if you prefer IDE-integrated AI features over the CLI.

Verdict

Twinny is a clever local-first bridge that fills a specific niche well, but at 25 stars and 0.80% credibility, it is early-stage software with limited community validation. Documentation is thorough and bilingual, tests are present, and the security posture includes configurable redaction and opt-out telemetry. However, the star count signals caution: this is a personal project with a narrow audience, not a battle-tested enterprise tool. Try it if your team lives in Lark and wants to steer Codex from chat.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.