FluxyChatFeatures
Ambient agents
Event-driven agents that react to webhooks and room events within policy bounds, not only @mention.
Ambient agents
Ambient agents run when external events or room conditions match a policy: a keyword in a message, a webhook payload, a schedule, and so on. They do not need an explicit @mention.
Console
Agents → Ambient (/agents/ambient)
Create policies with:
| Field | Meaning |
|---|---|
| Trigger | webhook, keyword, schedule, … |
| Max autonomy | How far the agent can act without human escalation |
| Target room | Where to post results |
| Agent ID | Which WorkflowAgent executes |
Enable or disable policies without deleting configuration.
Architecture
Webhook / room event → Worker policy engine (agent_policies)
→ WorkflowAgent → room message + audit logAPI
GET /admin/ambient-agents/policies
POST /admin/ambient-agents/policies
PATCH /admin/ambient-agents/policies/:id
DELETE /admin/ambient-agents/policies/:id
POST /webhooks/ambient/:policyIdGuardrails
- Scoped permissions per policy.
- Audit log on every autonomous action.
- Escalation above configured risk threshold.
Pair with tool approvals for sensitive tools.