AI Agents
The AI Agent Directory — register autonomous agents as governed identities with computed risk and trust scores.
The AI Agent Directory treats every autonomous agent as a first-class enterprise identity. Nothing autonomous runs anonymously.
The agent model#
An agent captures who owns it, what powers it, where it runs, and what it can reach:
| Field | Notes |
|---|---|
name, description, owner | Identity and accountability. |
businessUnit, department | Organizational ownership. |
modelProvider, modelVersion | e.g. Anthropic / claude-opus-4. |
environment | DEVELOPMENT · STAGING · PRODUCTION. |
lifecycleState | DISCOVERED → REGISTERED → APPROVED → DEPRECATED → RETIRED. |
authMethod | OAUTH · API_KEY · MTLS · NONE. |
businessCriticality | LOW · MEDIUM · HIGH · CRITICAL. |
| Capability flags | productionAccess, sensitiveDataAccess, internetAccess, externalToolUsage, canInvokeAgents. |
connectedSystems, connectedMcpCount | Blast-radius inputs. |
Risk & trust are computed#
You never send riskScore or trustScore. AuthSpoke computes them server-side from the agent's blast radius (what it can reach and do) and its governance posture (how it authenticates, its lifecycle). Scores are recomputed on every create and update. See Risk & Trust.
Register an agent#
curl -X POST "https://your-company.authspoke.com/api/v1/ai/agents" \
-H "Authorization: Bearer $TOKEN" -H "Content-Type: application/json" \
-d '{
"name": "support-copilot",
"owner": "[email protected]",
"modelProvider": "OpenAI",
"modelVersion": "gpt-4o",
"environment": "STAGING",
"authMethod": "API_KEY",
"internetAccess": true,
"externalToolUsage": true,
"connectedMcpCount": 1
}'
Lifecycle#
Move an agent through its lifecycle by updating lifecycleState. A DISCOVERED agent (found by Discovery) becomes governed when you promote it to REGISTERED or APPROVED. Each transition emits an audit event.
Everything is an event#
Create, update and delete all emit immutable events (AGENT_REGISTERED, AGENT_UPDATED, AGENT_DELETED) consumed by the live feed and per-agent timelines.
API#
See the AI Agents API for every endpoint, field, and example.