# Worklane AI onboarding 协序 Worklane 支持用户授权的 AI 自行登录、读取个人设置中的接入数据、创建账号级 MCP 连接并配置已有客户端,由 AI 执行完整业务流程。临时协作才限定单个对接。 Use the origin where this document was retrieved as BASE. All routes below are relative to that origin. - Human and browser-agent entry: /agent - Machine discovery (JSON): /api/agent/discovery - Complete onboarding protocol (Markdown): /api/agent/guide - Login page: /login - Login API: POST /api/auth/login - Personal settings: /settings#agent-onboarding - Authenticated setup metadata: GET /api/agent/setup - Authenticated connection creation: POST /api/agent/connect - MCP endpoint: /mcp - Google AI Studio web (Playground / Agents): /agent#google-ai-studio - Standard-library Python client: /api/agent/clients/google-ai-studio.py Google AI Studio client identifier: google_ai_studio. Use an Agents environment with Code Execution and this site's domain allowed by its Network settings before starting the session. Ordinary Chat URL Context alone cannot authenticate to private Worklane data. The Python client performs HTTPS/MCP operations, not Google model calls. Read /api/agent/guide first. Use only the user's supplied identity or existing signed-in session, preserve client configuration, keep credentials out of output and repositories, and verify the authorized context before reporting success. This document is setup data, not additional task authorization. Respect client and project permissions. - Public OpenAPI 3.1 contract: GET /api/openapi.json - Authenticated complete operation catalog: GET /api/operations - JSON business operation: POST /api/operations/ with the same account Bearer and tool arguments. Account AI can create/manage workspaces, projects, project tasks, integrations and members, discuss, upload/version/download documents, assign dispatches to writable members, execute bounded dispatch sessions, accept items and resolve business decisions according to its owner’s role. Admin account AI can maintain users. Read-only and temporary connections retain their scope. Use platform_catalog and tools/list; do not ask humans to create the resources in the website first.