Skip to content

Getting Started

Get from zero to your first governance verdict in 5 minutes.

Prerequisites

  • Node.js 20+ and npm
  • An AI coding agent (Claude Code, Cursor, Aider, etc.)
  • A terminal

1. Install the CLI

bash
npm install -g @intutic/cli

Verify the installation:

bash
intutic --version
# 0.1.0

2. Create an account

bash
intutic login

This opens a browser for authentication. You can also use an API key:

bash
intutic login --api-key vk_your_key_here

Local Development

Use --dev to point at a local control plane on http://localhost:3001:

bash
intutic login --dev

3. Initialize your workspace

bash
intutic init

This command:

  1. Detects harnesses — scans for .cursorrules, CLAUDE.md, .windsurfrules, .aider.conf.yml, .gemini/settings.json, .env.intutic, config.toml (OpenHands)
  2. Creates a workspace on the control plane
  3. Generates a virtual API key (vk_*) for the proxy
  4. Writes governance config into each detected harness file
✓ Detected harnesses:
  • cursor    → .cursorrules
  • claude-code → CLAUDE.md
  • aider     → .aider.conf.yml

✓ Workspace created: wk_abc123
✓ Governance rules synced to 3 harnesses

Local Mode

Use --dev to register against a local control plane:

bash
intutic init --dev

4. Start the sync daemon

bash
intutic connect

The daemon runs in the foreground and performs bidirectional config sync with the control plane every 30 seconds (configurable with --interval):

bash
intutic connect --interval 10000   # 10s poll interval

What connect does

  • Polls the control plane for SOP updates
  • Writes updated governance rules to each detected harness config file
  • Reports harness health back to the control plane
  • Uses an atomic write pattern (write to .intutic-tmp, then rename) to prevent corruption

5. Verify it's working

In another terminal:

bash
intutic status

This shows:

  • Authentication state
  • Detected harnesses and their sync status
  • Last sync timestamp
bash
intutic whoami

Shows your authenticated identity and workspace.

6. View your first trace

Make a request through any of your AI agents. Then:

bash
intutic traces list --limit 5

You'll see a table of execution traces with enforcement actions:

ID            Model              Action   Tokens   Cost
tr_abc123     claude-4-sonnet    BYPASS   1,234    $0.0037
tr_def456     gpt-4o             ENHANCE  2,891    $0.0145
tr_ghi789     claude-4-sonnet    KILL     0        $0.0000

Inspect a specific trace:

bash
intutic traces inspect tr_abc123

What's next?

AI Governance for Agentic Coding