Documentation8 pages
AI Workforce OS technical documentation
For anyone who wants to understand, deploy or integrate AI Workforce OS: how it is built, what the API exposes, how accounts are isolated and what can be configured. Everything describes what the code does today.
Concepts
How it is built and how the agents reason.
- AI Workforce OS Architecture Components, request flow, the WebSocket channel, data model, row-level-security multi-tenancy and how simulation differs from real AI.
- Agent runtime The Python agent runtime: HTTP contract, simulation vs live mode, model providers and fallback, data policy, tools, sandbox and cost budgets.
Integrate
The API, authentication and the skill format.
- API Reference Authentication, API tokens, errors and /api/v1 routes: requests, tasks, approvals, agents, projects, skills, shop, drive and the WebSocket.
- Authentication and multi-tenancy Accounts, sessions, two-factor auth, roles, invitations and Postgres row-level security: how AI Workforce OS isolates each organization.
- Skills format How to write a skill: a folder with SKILL.md, frontmatter, limits, review lifecycle, import and export, and how it reaches the model.
Operate
Deployment, configuration and the security model.
- Production deployment Deploy AI Workforce OS on your own server with Docker Compose: Caddy with automatic TLS, Postgres, Redis, backups, updates and rollback.
- Configuration reference Every AI Workforce OS environment variable with its default: server, authentication, budget, model providers and frontend build arguments.
- Security model How the damage an agent can do is bounded: credential vault, approvals, caps, kill switch, audit log, RLS and sandbox, with the current limits.
Explore more
Try it in your own office
Create your account and land in an office with your AI team. For now the AI runs in simulation mode: scripted responses, no cost and nothing to connect.