200+ Commands That Let AI Agents Directly Control Lark (Feishu)

Overview

Lark (Feishu) has released an official open‑source command‑line tool called lark-cli. It covers 11 core business areas, provides more than 200 commands and 19 AI Agent skills, and is designed from day one to treat AI agents as first‑class citizens.

Why a Lark CLI?

Typical GUI actions such as checking calendars, sending messages, editing documents, managing tasks, or reading mail require multiple clicks. Directly using Lark’s OpenAPI is possible but involves handling authentication, pagination, and parameter formatting. lark-cli encapsulates all of this, allowing a single command to replace dozens of lines of code.

Business Domain Coverage

Calendar – view, create meetings, manage attendees, query free slots

Message – send/reply, group management, view history, download media

Cloud Docs – create/edit/search, Markdown support, whiteboard integration

Drive – upload/download, document search, comment management

Sheets – spreadsheet CRUD, data export

Base – multi‑dimensional tables, field configuration, record operations, data analysis

Task – create/update/complete tasks, subtasks, reminders, team assignment

Wiki – knowledge space and node management

Contacts – search by name, email, phone

Mail – browse/search, compose, draft management

Meeting – recording search, minutes with summary and transcription

Three‑Layer Architecture

The command design follows three layers:

Shortcut commands (prefixed with +) provide intelligent defaults, table output, and a dry‑run preview. Example: lark-cli calendar +agenda shows today’s agenda.

API commands – over 100 commands auto‑generated from OpenAPI metadata, each mapped one‑to‑one with platform endpoints and validated for quality.

Raw API access – direct invocation of more than 2,500 Lark API endpoints with custom parameters and request bodies, covering any conceivable operation.

AI Agent‑Native Design

The tool ships with 19 structured Agent Skills that AI models such as Claude Code can call directly. These skills expose core Lark capabilities like calendar, instant messaging, docs, drive, sheets, base, tasks, mail, contacts, wiki, events, video conference, whiteboard, meeting minutes, API explorer, custom skill maker, meeting summary, and stand‑up report.

After installation, an AI Agent can be instructed with natural language, e.g., “show me today’s calendar”, “send a message to Alice”, or “export this document as a sheet”, and lark-cli will execute the appropriate command.

Getting Started

Installation and configuration are straightforward:

npm install -g @larksuite/cli

lark-cli config init

lark-cli auth login --recommend

For Claude Code integration, an extra step is required:

npx skills add larksuite/cli -y -g

Practical Use Cases

Scenario 1: Quickly view today’s agenda with lark-cli calendar +agenda, producing a formatted table faster than opening the app.

Scenario 2: Send a message from the terminal:

lark-cli im +messages‑send --chat-id "oc_xxx" --text "Report completed"

.

Scenario 3: Use the lark‑workflow‑meeting‑summary skill to automatically search, aggregate, and generate meeting minutes.

Scenario 4: Automate multi‑dimensional table operations via the lark‑base skill—create tables, configure fields, ingest data, and build board views.

Rich Output Formats

lark-cli

supports several output options to suit different pipelines: --format json – raw JSON for programmatic consumption --format pretty – human‑readable pretty‑printed output --format table – tabular view --format ndjson – newline‑delimited JSON for streaming --format csv – CSV for spreadsheet import

Advanced flags include --dry-run for preview, --page-all for automatic pagination, and schema to query API parameter structures.

Security Considerations

Because AI agents act on your Lark account, security is paramount. lark-cli implements:

Input injection protection to block malicious prompts

Terminal output sanitization to remove sensitive data

OS‑native keychain storage for credentials

Fine‑grained permission scope selection at login

The project warns that agents operate within the granted scopes and may expose sensitive data; it recommends keeping default security settings, limiting bot access to private conversations, and monitoring authorized scopes.

Why It Matters

This is the first Agent‑native CLI among major Chinese SaaS platforms, signaling a broader trend where SaaS products prioritize AI Agent interaction over traditional GUIs. For developers, it opens possibilities such as cross‑tool automation workflows, AI‑driven Lark assistants, CI/CD integration for notifications and document updates, and custom skill development via lark‑skill‑maker.