AgentFlow documentation
Human-in-the-loop approval authority for AI agents. Integrate the agent API or run the operator console.
AgentFlow sits between AI agents and the systems they act on. An agent submits an authority request describing the action it wants to take; a human approver reviews it in the TrustLoop app, verifies their identity, and records a decision; AgentFlow finalizes the request and issues a signed Ed25519 receipt the agent can poll or receive by webhook. Every request, decision, and proof is scoped to a workspace and written to a tamper-evident audit trail.
Quickstart
Get an agent key and run the full request → decision → proof loop with curl.
AI agents
Submit authority requests with x-ai-agent-key, poll status, and fetch signed proofs.
Using the platform
Operate the console: agent connections, routing, people, and audit.
API reference
Generated reference for every agent API procedure.
Who these docs are for
- AI-agent integrators — you call the agent API to submit requests and collect receipts. Start with the quickstart then AI agents.
- Operators — you run the console: create agent connections, configure routing, invite approvers, and review the audit trail. See Using the platform.
Platform endpoints
| Surface | URL |
|---|---|
| Agent API | https://af-api.ivalt.com/api/v1 |
| Console | https://af-dashboard.ivalt.com |