Skip to main content

Quick Start

Create your first governance policy and deploy it in under 5 minutes.

Prerequisites​

  • Metatate installed and running (Installation)
  • At least one table reference configured via DATABASE_CATALOG
  • A warehouse configured via APP_WAREHOUSE

Step 1: Open the Policy Editor​

Navigate to the Policies page from the left sidebar. Click New Policy to open the policy editor.

Step 2: Describe Your Policy​

In the natural language input, describe what you want to govern. For example:

"Protect customer PII in the CUSTOMERS table. Email and phone should be masked for non-admin roles. The table contains GDPR-regulated data and should only be used for customer support and analytics."

Click Generate — Metatate's AI (powered by Snowflake Cortex) will generate a structured YAML policy.

Step 3: Review the Generated Policy​

The editor shows:

  • YAML view — the full policy definition
  • Visual review — a structured breakdown of each instruction

Review the generated policy. You can edit the YAML directly or regenerate with different instructions.

Step 4: Save and Deploy​

  1. Click Save to store the policy
  2. Navigate to the Deployments page
  3. Select your policy and click Plan — the diff engine compares desired vs. deployed state
  4. Review the deployment plan (add/update/remove actions)
  5. Click Apply to deploy

Step 5: Verify​

After deployment, the governance materializer populates 6 governance tables. Verify by:

  • Checking the Catalog page — your table should show governance metadata
  • Testing MCP tools — go to MCP Tools → Test Tools tab and run discover-context

What Just Happened?​

Your policy went through the full governance pipeline:

  1. AI Generation — Natural language → structured YAML policy
  2. Instruction Extraction — Policy → typed instructions (classification, masking, usage, etc.)
  3. Compilation — Diff engine compared desired state vs. deployed state
  4. Deployment — Policy state persisted to Snowflake
  5. Materialization — 6 governance tables rebuilt via atomic swap
  6. MCP — Governance data now available to AI agents via MCP tools

Next Steps​