Agents
Register, configure, and govern AI agents in your workspace.
An agent represents any AI system that takes actions — chatbots, research assistants, ops bots, or custom workflows. Agents are registered per workspace with identity, provider, environment, and optional allowlists.
Agent fields
| Field | Values / notes |
|---|---|
| name | Required. 2–100 characters |
| provider | Required. openai, anthropic, gemini, openrouter, custom |
| status | active, inactive, pending (default), archived |
| environment | production, staging, development (default) |
| risk_level | low, medium (default), high, critical |
| slug | URL-safe identifier, auto-generated if omitted |
| allowed_tools[] | MCP/external tools this agent may invoke |
| allowed_resources[] | Resources this agent may access |
| api_key | Optional agent secret — stored as SHA-256 hash + last4 |
Lifecycle
- POST /agents — register (emits agent.created webhook)
- PATCH /agents/{id} — update metadata, allowlists, or agent key
- GET /agents — list with cursor pagination, filter by status and provider
- DELETE /agents/{id} — remove from workspace