Deployments
The Deployments page is where policies move from draft to production. Metatate uses a plan-and-apply workflow inspired by infrastructure-as-code tools -- you always see exactly what will change before anything is applied.
the deployments page
Deployment Workflow
1. Select Policies
Choose which policies to include in the deployment. You can deploy one policy at a time or batch multiple policies together. The page shows each policy's current status (draft, modified, or unchanged).
2. Plan
Click Plan to generate a deployment plan. The diff engine compares your desired policy state against what is currently deployed and produces a detailed diff.
a deployment plan with adds, updates, and removes
The plan shows three categories of changes:
| Change Type | Description |
|---|---|
| Add | New policies or instructions that will be created |
| Update | Existing policies where the definition has changed |
| Remove | Policies or instructions that will be deleted |
Review the plan carefully. No changes are made to governance tables at this stage.
3. Apply
Click Apply to execute the plan. The policy manager processes each change:
- New policies are inserted into the deployed policy store.
- Updated policies are replaced with their new versions.
- Removed policies are deleted from the deployed state.
the apply progress/confirmation
4. Materialize
After applying, the materializer rebuilds the six governance tables that power MCP tools and downstream queries. The materialization process uses an atomic write strategy for reliability:
- Write new data to staging tables.
- Validate the staging tables.
- Perform an atomic swap from staging to production.
This ensures that MCP tools always read a consistent, complete snapshot -- never a partial update.
Governance Tables
The materialization step produces these tables in the app_data schema:
| Table | Contents |
|---|---|
governance_catalog | All tables and columns under governance |
governance_policies | Deployed policy definitions |
governance_classifications | Column-level classification labels |
governance_masking_rules | Masking strategies and role exceptions |
governance_usage_guidance | Usage guidance for tables and columns |
governance_access_rules | Role-based access control rules |
After Deployment
Once materialized, the governance data is immediately available to:
- MCP Tools -- AI assistants can query governance rules via the MCP endpoint.
- Activity Log -- A
deploy.appliedentry is recorded in the Activity Log. - Dashboard -- Stats on the Dashboard update to reflect the new state.
Tips
- Always review the plan before applying. The diff view exists to prevent accidental changes.
- If a deployment looks wrong, create corrective policies and redeploy -- the next plan will show the fixes.
- Materialization is idempotent. Running it again with the same policies produces the same result.