The Enterprise CLI That Treats AI Agents as First-Class Users

What it does lark-cli is the official terminal interface for Lark (Feishu), ByteDance’s enterprise collaboration platform. It bundles more than 200 commands covering chat, documents, spreadsheets, mail, calendar, and HR workflows into a single tool. The project is written in Go and published under an MIT license by the larksuite team.

The interesting bit The unusual part is its agent-native design: it ships 26 structured Skills—pre-tested command sets with concise parameters and structured output—so popular AI agents can operate the platform with zero extra setup. A three-layer architecture lets humans use shortcuts while agents call curated API commands or drop down to raw OpenAPI when needed.

Key highlights

• Covers 18 business domains, from Messenger and Base to OKR and Attendance.
• 26 AI Agent Skills with identity switching between user and bot contexts.
• Security guardrails include input injection protection, output sanitization, and OS-native keychain storage.
• OAuth flows support non-blocking agent authentication via device codes and polling.
• Commands are tested against real agents to maximize call success rates.

Caveats

• Despite being written in Go, the distribution depends on a Node.js runtime, and building from source also requires Python 3—a somewhat baroque toolchain.
• Project-management features via Meegle are split into a separate CLI rather than bundled here.
• The README alternates between citing 24 and 26 agent skills, so the exact inventory is slightly inconsistent.

Verdict Worth a look if you are automating Lark workflows or integrating an AI agent into an enterprise stack. Skip it if you are not using Lark/Feishu; the tool is tightly coupled to that ecosystem.