One Open Source Project a Day (Part 62): lark-cli - Official Lark/Feishu CLI & AI Agent Skills

Introduction

"Built for humans and AI Agents."

This is Part 62 of the "One Open Source Project a Day" series. Today's project is lark-cli (GitHub).

Need to drive Lark/Feishu from the terminal or an AI workflow? lark-cli is the official CLI from the larksuite team: 11 business domains, 200+ commands, 19 AI Agent Skills, and a three-layer command model (Shortcuts → API Commands → Raw API). Parameters and structured outputs are tuned for agents; OAuth login, user/bot switching (--as), dry-run, and security-conscious defaults. MIT licensed; install with npm install -g @larksuite/cli.

What You'll Learn

Positioning: humans + agents
Three-layer commands and typical usage
What the 19 Agent Skills cover
Auth, scopes, --as user/bot
Security notes and boundaries

Prerequisites

Basic understanding of Feishu/Lark tenant and Open Platform apps
Terminal and npm familiarity (optional)

Project Background

Project Overview

lark-cli is the official command-line tool for Lark / Feishu, calling Open Platform APIs for calendar, IM, cloud docs, Drive, Base (Bitable), Sheets, tasks, wiki, contacts, mail, meetings, and more. It is Agent-Native: ships structured Skills compatible with Cursor, Claude, and similar tools to reduce ad-hoc API wiring.

Maintainer

Team: larksuite (official)
Implementation: Go (source builds need Go 1.23+); primary distribution via npm @larksuite/cli
Companion: install CLI SKILL with npx skills add larksuite/cli -y -g

Project Data

⭐ GitHub Stars: ~6k+ (see repo)
📄 License: MIT
📦 npm: @larksuite/cli
🌐 Product: Lark / Feishu Open Platform

Core Features

Main Role

One tool spans Messenger, Docs, Drive, Base, Sheets, Calendar, Mail, Tasks, Wiki, Contact, Meetings; humans use + shortcuts, power users use API commands aligned with platform metadata, and fallback uses lark-cli api for 2500+ Open APIs. Output formats: json / pretty / table / ndjson / csv for scripts and agents.

Use Cases

Calendar & meetings: calendar +agenda, create events, free/busy; meeting minutes flows (with relevant skills)
IM & groups: send/search messages, members, file transfer
Docs & sheets: Markdown docs, cell read/write, export
Base (Bitable): tables, fields, records, views (with lark-base skill)
Mail & tasks: compose/send mail, todos and lists

Quick Start

npm install -g @larksuite/cli
npx skills add larksuite/cli -y -g
 
lark-cli config init
lark-cli auth login --recommend
lark-cli auth status
 
lark-cli calendar +agenda

Build from source (Go 1.23+, Python 3):

git clone https://github.com/larksuite/cli.git && cd cli
make install
npx skills add larksuite/cli -y -g

Core Capabilities

Three layers: + shortcuts → domain API subcommands → lark-cli api raw HTTP
19 Agent Skills: e.g. lark-calendar, lark-im, lark-doc, lark-base, lark-mail, lark-vc
Identity: --as user / --as bot under the same app
Schema: lark-cli schema for params, body, scopes
Security defaults: input safety, output sanitization, OS keychain storage (see official warnings)
Pagination & dry-run: --page-all, --dry-run

Project Advantages

Comparison	lark-cli	Hand-rolled curl/SDK	Web UI only
Onboarding	One CLI + OAuth	Manual doc diving	No code
Agents	Skills + structured output	Custom wrappers	Not automatable
Coverage	200+ commands + API escape hatch	You maintain it	UI-limited
License	MIT OSS	Varies	—

Why this project?

Official maintenance and OpenAPI alignment
Serves both humans and agents
MIT-friendly for internal tooling

Deep Dive

Three-Layer Command System

Layer	Form	Purpose
Shortcuts	`lark-cli <domain> +<action>`	Human & agent friendly defaults
API commands	e.g. `lark-cli calendar calendars list`	1:1 curated platform mappings
Raw API	`lark-cli api GET	POST ...`

Agent Skills (sample)

Skill	Summary
`lark-shared`	Config, login, identity, scopes, security
`lark-calendar`	Events, agenda, free/busy
`lark-im`	Messages, groups, search, media
`lark-doc` / `lark-drive` / `lark-sheets` / `lark-base`	Docs, files, sheets, Base
`lark-mail` / `lark-task` / `lark-contact` / `lark-wiki`	Mail, tasks, directory, wiki
`lark-event` / `lark-vc` / `lark-minutes`	Events WS, meetings, minutes
`lark-openapi-explorer`	Explore underlying APIs from docs
`lark-workflow-*`	Meeting summaries, standup-style reports

Full table: README - Agent Skills.

Auth & Agent Flow

lark-cli config init / config init --new: app credentials
lark-cli auth login --recommend: common scopes
Agent-friendly: auth login --no-wait for URLs; resume with device-code polling

Security & Risk

The CLI can be driven by AI; risks include hallucination, injection, and over-privileged actions. After authorization, the agent acts within granted scopes as you. Do not expose integrated bots to untrusted group chats; avoid weakening default protections. Read Security & Risk Warnings before use.

Project Links and Resources

Official Resources

🌟 GitHub: https://github.com/larksuite/cli
📦 npm: https://www.npmjs.com/package/@larksuite/cli
📚 中文 README: README.zh.md
🐛 Issues: https://github.com/larksuite/cli/issues

Target Audience

Developers automating Feishu/Lark
Teams wiring Lark capabilities into Cursor, Claude, or similar agents
Users who want official semantics without heavy glue code

Visit my homepage for more useful knowledge and interesting products