Ctrl AI

Quickstart

Create your first organization, unit, and run a query in 5 minutes.

Prerequisites

  • A Ctrl AI account (cloud) or a running self-hosted instance
  • Admin access to create an organization

1. Create an Organization

After signing in, you'll be prompted to create your first organization. This is your team's workspace — all units, workflows, and audit logs are scoped to it.

Choose a name and slug (URL-friendly identifier). You can configure SSO, billing, and LLM providers later in Settings.

2. Create a Domain

Navigate to Settings > Domains and create your first knowledge domain. Domains organize your units by department or expertise area (e.g., "Finance", "Legal", "HR").

3. Create Your First Unit

There are several ways to create units. The fastest is AI-assisted generation:

  1. Go to Build in the sidebar
  2. Select your domain
  3. Type a question that represents the kind of reasoning you want to capture
  4. The AI generates 3-5 candidate units with full formal definitions
  5. Review and approve the ones that look correct

The approved units enter the verification queue at /review.

4. Verify the Unit

Navigate to Review in the sidebar. You'll see your candidate units waiting for verification.

As a reviewer, you can:

  • Approve individual elements (steps, criteria, conditions)
  • Suggest edits for elements that need adjustment
  • Approve the unit when all elements look correct

A unit needs ≥80% approval rate from ≥2 reviewers to reach "verified" status.

The unit creator cannot be the sole verifier — this enforces separation of duties for audit compliance.

5. Run a Query

Go to Ask in the sidebar and type a question related to your domain. The inference engine will:

  1. Parse your query into typed variables
  2. Find matching verified units
  3. Execute them (deterministic programs + constrained LLM reasoning)
  4. Return a trust-tagged response with per-claim attribution

Each segment of the response is color-coded by trust level:

  • Green — deterministic, program-computed
  • Blue — expert-reviewed structured reasoning
  • Yellow — synthesized but not yet reviewed
  • Gray — neural-only (no unit coverage)

6. View the Audit Trail

Every query is logged with a full execution trace. Go to Audit in the sidebar to see:

  • Which units fired and their verification status
  • What inputs were used and what outputs were produced
  • Trust level summary and coverage percentage
  • Which LLM models were used (model provenance)

Next Steps

On this page