lark

A CLI tool for interacting with Lark/Feishu APIs, designed for use with Claude Code and other AI assistants.

Features

• Calendar - List, create, update, delete events; check availability; find common free time; RSVP
• Contacts - Look up users by ID, search by name, list department members
• Documents - Read documents as markdown, list folders, resolve wiki nodes, get comments
• Messages - Retrieve chat history, download attachments, send messages, add/list/remove reactions
• Mail - Read and search emails via IMAP with local caching
• Minutes - Get meeting recording metadata, export transcripts, download media

Fork Differences

This is a fork of yjwong/lark-cli with the following changes:

Feishu (China) Region Support

The original project only supports Lark (international). This fork adds support for Feishu (China) by setting region: "feishu" in config or LARK_REGION=feishu as environment variable. This switches all OAuth and API endpoints from larksuite.com to feishu.cn.

Reduced OAuth Scopes

Some scopes require company admin approval on Feishu, which blocks normal login. This fork removes those scopes so you can authenticate without waiting for admin approval. The trade-off is that the following features are unavailable:

Removed Scope	Impact
contact:department.organize:readonly	Cannot view department org structure
docs:doc:readonly	Cannot read legacy Docs (new Docx still works)
drive:drive:readonly	Cannot browse Drive root folder
space:document:retrieve	Cannot retrieve wiki space documents
mail:* (4 scopes)	Mail commands do not work
minutes:* (2 scopes)	Minutes commands do not work

Additional Features

For a full list of enhancements added in this fork beyond the above, see the GitHub Issues.

Why This Tool?

The official Lark MCP server exists, but its tools are not token-efficient. Each tool call returns verbose responses that consume significant context window space when used with AI assistants.

This CLI addresses that by:

• Returning compact JSON - Structured output optimized for programmatic consumption
• Providing markdown conversion - Documents are converted to markdown (~2-3x smaller than raw block structures)
• Supporting selective queries - Fetch only what you need (e.g., just event IDs, just document titles)

The result: AI assistants can interact with Lark using fewer tokens, leaving more context for actual work.

Quick Start

1. Create an app at the Developer Console:
   • Lark: https://open.larksuite.com
   • Feishu: https://open.feishu.cn
2. Copy config.example.yaml to .lark/config.yaml and add your App ID
3. Set LARK_APP_SECRET environment variable
4. (Feishu users) Set LARK_REGION=feishu or add region: "feishu" to config
5. Run ./lark auth login to authenticate
6. Start using: ./lark cal list --week

See USAGE.md for full documentation.

Building

make build    # Build binary to ./lark
make test     # Run tests
make install  # Install to $GOPATH/bin

Usage with Claude Code

This tool is designed to be invoked via Claude Code skills. Pre-built skill definitions are included in the skills/ directory.

Installing Skills

Copy the skill directories to your Claude Code skills location:

# Project-specific (recommended)
cp -r skills/* /path/to/your/project/.claude/skills/

# Or user-wide
cp -r skills/* ~/.claude/skills/

Available skills:

• calendar - Manage calendar events, check availability, RSVP
• contacts - Look up users and departments
• documents - Read documents, list folders, browse wikis
• messages - Retrieve chat history, download attachments, send messages to users and chats
• email - Read and search emails via IMAP with local caching
• minutes - Get meeting recordings, export transcripts, download media

Configuration

The skills assume lark is in your PATH. If not, you can either:

1. Add the binary location to your PATH
2. Edit the skill files to use the full path
3. Set LARK_CONFIG_DIR environment variable to point to your .lark/ config directory

The JSON output format makes it straightforward for AI assistants to parse responses and take action.

License

MIT - see LICENSE