Lark CLI (lark-cli) is the official Lark/Feishu command line, maintained by the larksuite team for people and AI agents. The repo larksuite/cli is Go, MIT, created 2026-03-25. On 2026-09-20 GitHub listed 17,312 stars and 1,390 forks. Those counts are a heat signal, not traffic. The GitHub description says 200+ commands and 20+ AI Agent Skills across Messenger, Docs, Base, Sheets, Calendar, Mail, Tasks, Meetings, and more. The README also says 24 structured Skills in one place and 26 in another. Treat the range as first-party wording, not a measured inventory.
Install from npm as @larksuite/cli. Compare Claude Code or Codex CLI if you needed a coding agent rather than a Feishu operator, and Agent Skills if you wanted a broader skill pack instead of Lark's own skills.
Key Features
- Agent-native skills: bundled Skills for calendar, IM, docs, drive, markdown, sheets, slides, Base, tasks, mail, contacts, wiki, events, meetings, whiteboard, OpenAPI explorer, skill maker, attendance, approval, OKR, and a few workflows.
npx skills add larksuite/cli -y -ginstalls the skill pack. - Three command layers:
+shortcuts for humans and agents, curated API commands mapped to Open Platform endpoints, andlark-cli apiraw calls for the rest of the surface. - Auth for agents:
lark-cli config initandlark-cli auth login --recommend.--no-waitreturns a verification URL so an agent can hand the browser step to you. - JSON contract:
--format jsonwrites{ "ok": true, ... }to stdout. Errors go to stderr withok: false. Checkok, not a raw OpenAPIcode. - Enterprise embed path: wrapper
mainunderextension/plus embed docs. Append.mdto Open Platform doc URLs for raw Markdown.
Limitation: the README's security section is explicit. An agent with your OAuth scopes can leak data or take actions you did not intend. Default protections exist; relaxing them raises risk. Do not drop the bot into group chats. Risk-control signals (OS type and hardware model) go to official Feishu/Lark HTTPS domains unless you turn them off.
Use Cases
- Individuals who want agenda, docs, and chat from a terminal or a coding agent.
- IT / ISV embedding Feishu into an internal agent with centralized credentials.
- Skip if you only needed a general coding CLI, or if you cannot accept Feishu OAuth on the machine.
Pricing
| Piece | Price | Notes from first-party pages 2026-09-20 |
|---|---|---|
| CLI | $0 | MIT. npx @larksuite/cli@latest install. About 17,312 stars. |
| Feishu/Lark seat | Your existing plan | The CLI calls Open Platform APIs under your app and user. |
| Enterprise embed | Your infra | Wrapper path; no separate CLI SKU in the README. |
Getting Started
npx @larksuite/cli@latest installlark-cli config initthenlark-cli auth login --recommendlark-cli auth statusthenlark-cli calendar +agenda
First-party start: README and the embed guide.
Frequently Asked Questions
Feishu or Lark?
Same product family. Feishu is the China brand; Lark is the international brand. The CLI covers both Open Platforms.
Why do people complain about too many skills?
The pack is wide on purpose. Some users load a router skill instead of every domain skill. That is a preference, not a defect in the CLI.
Can an agent install it without me?
Install and status checks yes. config init and auth login still need you in a browser.
Alternatives
- Claude Code: coding agent that can call Lark CLI as a tool.
- Codex CLI: OpenAI terminal agent, not a Feishu client.
- n8n: workflow automation if you wanted a GUI instead of a CLI.
Tips
- Start with
--recommendscopes. Add domains later withauth login --domain. - Use
--dry-runon send/create shortcuts before you let an agent loose. - Prefer shortcuts over raw
apiuntil you know the endpoint.
Conclusion
Lark CLI is the official way to let a human or an agent operate Feishu from a terminal. Install it if that is the surface you already live in. Keep the default security settings, and read the risk section before you widen scopes.
Comments
No comments yet. Be the first to comment!
Related Tools
Related Insights
Skills + Hooks + Plugins: How Anthropic Redefined AI Coding Tool Extensibility
An in-depth analysis of Claude Code's trinity architecture of Skills, Hooks, and Plugins. Explore why this design is more advanced than GitHub Copilot and Cursor, and how it redefines AI coding tool extensibility through open standards.
Stop Cramming AI Assistants into Chat Boxes: Clawdbot Picked the Wrong Battlefield
Clawdbot is convenient, but putting it inside Slack or Discord was the wrong design choice from day one. Chat tools are not for operating tasks, and AI isn't for chatting.
What Locks Codex Is the Picker, Not the Models
You already pay for OpenCode Go, Grok, and Z.ai, but the Codex picker still shows mostly GPT. The community project codex-router does not teach another install ritual. It puts subscriptions you already bought back into the selector, keeps keys on the machine, and leaves native GPT alone.