Purpose
Lark/飞书 CLI is the official command-line tool maintained by the larksuite team, designed for two types of users: human users and AI Agents. It encapsulates the business capabilities of the Lark Open Platform as command-line operations, covering core business domains such as Messenger, Docs, Base, Sheets, Slides, Calendar, Mail, Tasks, Meetings, Markdown, and more, with a total of 18 business domains, 200+ curated commands, and 26 AI Agent Skills.
Main Capabilities
- Agent-native design: Provides 24 structured Skills out of the box, compatible with mainstream AI tools, so Agents can operate Lark without additional configuration.
- Broad coverage: 18 business domains, 200+ curated commands, 26 AI Agent Skills.
- Optimized for AI: Every command has been tested with real Agents, with concise parameters, smart defaults, and structured output to improve Agent invocation success rates.
- Three-tier command system: Shortcut commands (prefixed with +, for humans and AI, with smart defaults, table output, and dry-run preview) → API commands (automatically generated from Lark OAPI metadata, filtered through evaluation and quality gates, with 100+ commands corresponding one-to-one with platform endpoints) → raw API calls (can directly call any Lark Open Platform endpoint, covering 2500+ APIs).
- JSON output contract: When using --format json (the default), success and error envelopes are distinguished from each other. Success is written to stdout with exit code 0; errors are written to stderr with a non-zero exit code. To determine whether a command succeeded, check ok == true (or the exit code), not code == 0.
- Schema introspection: Use schema to view the parameters, request body, response structure, supported identities, and permission scopes of any API method.
- Security and controllability: Includes input injection protection, terminal output sanitization, and OS-native keychain credential storage.
Usage
Prerequisites: Node.js (npm/npx); when building from source, Go v1.23+ and Python 3 are also required.
Quick start for human users (choose one of two):
- Option 1 (recommended): Install from npm.
- Method 2: Build from source (requires Go v1.23+ and Python 3).
AI Agent Quick Start:
- Step 1: Install.
- Step 2: Configure application credentials. Run the command in the background; it will output an authorization URL, which you should extract and send to the user; after the user completes setup in the browser, the command exits automatically.
- Step 3: Same as above, run in the background, extract the authorization URL and send it to the user.
- Step 4: Verify.
Other key points:
- Run lark-cli <service> --help to view all shortcut commands.
- For commands with side effects, you can first preview the request with --dry-run.
- Tip for AI Agents: Append .md to any Open Platform documentation URL to get the raw Markdown.
Use Cases
Suitable for scenarios that require automating Feishu/Lark operations via the command line or an AI Agent, such as operations and integration across business domains including messages, documents, Base, sheets, slides, calendars, email, tasks, and meetings.
Limitations and Risks
This tool can be invoked by an AI Agent to automate operations on the Feishu Open Platform, and there are inherent risks such as model hallucination, unpredictable execution, and prompt injection. After authorizing Feishu permissions, the AI Agent will act as your user identity within the authorized scope, which may lead to high-risk consequences such as sensitive data leakage or unauthorized operations, so please use it with caution. The tool enables security protections at multiple layers by default, but risks still exist; it is strongly recommended not to proactively modify any default security settings. Once the relevant restrictions are relaxed, the risk will increase significantly, and the consequences shall be borne by the user. It is recommended to use the Feishu bot integrated with this tool as a private conversation assistant, and not to add it to group chats or allow other users to interact with it, in order to avoid permission abuse or data leakage.
To reduce the security risk of access token theft, the CLI sends a minimized risk-control signal when making OpenAPI requests to the exact official Feishu/Lark HTTPS domains, to help identify abnormal API activity. This protection is enabled by default, and the information sent is limited to: operating system type (macOS, Windows, or Linux) and device hardware model (for example, Mac17,9). Commands can be used to disable or enable this protection for the current workspace or restore the default policy.
By using this tool, you are deemed to voluntarily assume all related responsibility. The project is open source under the MIT license; at runtime it calls the Feishu/Lark Open Platform APIs, and use of these APIs requires compliance withFeishu User Terms of Service, Feishu Privacy Policy, Feishu Open Platform Application Service Provider Security Management Specifications, Lark User Terms of Service, and Lark Privacy Policy.
Contributing
Community contributions are welcome. If you find a bug or have a feature suggestion, you can submit an Issue or Pull Request; for major changes, it is recommended to discuss via an Issue first. Before submitting a PR, please refer to AGENTS.md for the local build, test, and PR checklist.
