Lark CLI is a command-line tool for the Lark and Feishu workplace suite, built for both people and AI agents. It covers 18 business domains, including Messenger, Docs, Base, Sheets, Slides, Calendar, Mail, Tasks, Meetings, Wiki, Approval, and OKR, through more than 200 curated commands, and ships 26 agent skills so an agent can operate Lark without extra setup.
Commands come in three layers. Shortcuts (prefixed with +) wrap common operations with defaults, table output, and dry-run previews. API commands are generated from Lark OpenAPI metadata and map one-to-one to platform endpoints. Raw API calls reach any of the more than 2,500 Lark Open Platform endpoints directly.
Output defaults to JSON with separate success and error envelopes, and also comes as pretty text, tables, NDJSON, or CSV. Commands run as either the signed-in user or a bot. Organizations that embed the CLI in their own agent or platform extend it through a wrapper with centralized credentials, audit logging, and a restricted command surface, without changing the CLI source.
Features
- Agent skills: 26 skills covering calendar, messaging, docs, drive, sheets, slides, base, tasks, mail, wiki, meetings, approvals, OKRs, events, and workflow reports such as meeting summaries and standups
- Three command layers: shortcuts, API commands mapped 1:1 to platform endpoints, and raw calls to any Open Platform endpoint
- Domain coverage: Messenger, Docs, Drive, Markdown, Base, Sheets, Slides, Calendar, Tasks, Wiki, Contact, Mail, Meetings, Attendance, Approval, OKR, and Apps
- Structured output: JSON success and error envelopes on separate streams, plus pretty, table, NDJSON, and CSV formats
- Agent-friendly login: OAuth login with scope selection by domain, a non-blocking mode that returns a verification URL, and identity switching between user and bot
- Schema introspection: inspects any API method's parameters, request and response structure, identities, and scopes
- Dry run and pagination: previews side-effecting requests and pages through results automatically
- Security layers: input injection protection, terminal output sanitization, OS keychain credential storage, and DPoP token binding
- Event subscriptions: real-time WebSocket events with regex routing in an agent-friendly format
