Introduction
What Minctrl is, and the core concepts behind governed AI workflows.
Minctrl is the AI-native workflow builder for regulated operations. You design a process once; AI agents run it end-to-end; and a governance layer keeps a human on exactly the steps that carry real risk.
The problem it solves: automating a regulated process (prior authorization, KYC, claims, sanctions screening) usually forces a bad trade — either a human touches every step (slow, expensive) or agents run unsupervised (fast, uninsurable). Minctrl splits the difference. Agents do the work; a risk model decides, per step, whether the run can proceed on its own or must park for a human verdict.
Core concepts
Governed flows
A process template (a "vertical") is a canvas graph of steps. Each step is decorated with a risk tier and, where needed, a gate. The same graph drives both execution and governance.
Gates
A gate is a step where the run pauses for a human. When a run reaches a gate it
parks; a person records a verdict (approved / denied) against the gate's
step-id, and the run resumes from exactly where it stopped.
Connectors
A connector binds a step's abstract tool-id to a real HTTP API — your screening service, EHR, or ledger. Tools that do something irreversible stay shadowed (dry-run) until you explicitly opt in.
Durable park/resume
Runs are durable. A parked run can wait minutes or days; state is persisted, so resuming is a single API call — nothing is recomputed and no context is lost.
How the pieces fit
- Templates — start from one of the governed flow templates (
/process-templates). - Connectors — bind each step's tool-id to your real API (
/connectors). - Runs — start a governed run; it executes until it hits a gate, then parks.
- Resume — a human signs the gate;
/{id}/resumerecords the verdict and continues.
Every resource is scoped to your company — you only ever see your own data. See Multi-tenancy for how isolation works.
Ready to run one end-to-end? Head to the Quickstart.