Naïve Primitives — one API call for every real-world capability.

Trust & Ops

Agent Guardrails

Why all of the above is safe to run — human approval before risky actions, secrets it uses but never reads, expiring access, and every action audited.

AgentApprovals

Human-in-the-loop enforcement at the primitive boundary.

AgentVault

Per-user encrypted credentials — retrieved at runtime, never in prompts.

AgentSessions

Revocable MCP sessions — identity and policy travel with every call.

AgentLogs

Per-user activity audit trail.

AgentJobs

Track every async job — renders, formations, deploys — in one API.

AgentWebhooks

Real-time events from every primitive, pushed to your backend.