Unity AI Gateway observability

Unity AI Gateway observability lets you monitor cost, usage, performance, and access for every AI service from one control plane, backed by Unity Catalog system tables. As teams adopt coding agents, copilots, and custom agents across the organization, it gets harder to see who accesses which models or tools, how much they cost, and whether endpoints are healthy. Because every request routes through Unity AI Gateway, you can answer those questions in one place.

Note

Access to these system tables requires both account and metastore admin roles by default. Admins can manage access to system tables or publish the AI Gateway usage dashboard to share usage analysis without granting table access.

Requirements

Monitor cost

From the workspace UI, go to AI Gateway > Govern > Usage Dashboard > Cost Analysis to track dollar spend by endpoint, Azure Databricks-provided foundation model, or user over time. The built-in dashboard draws on usage and billing tables:

Note

External model spend is calculated from each provider's published prices and is informational only. See External models.

To attribute spend along custom dimensions, use request or service tags to group cost by business unit, project, or any dimension your organization cares about (see Track foundation model spend by user, team, or project).

See Monitor Unity AI Gateway cost.

Monitor usage and performance

The Unity AI Gateway usage dashboard (AI Gateway > Govern > Usage Dashboard) tracks request volume, token consumption, and latency for all endpoints governed through the gateway. Each Model or MCP Service also has a built-in Metrics tab for health monitoring without leaving the Catalog Explorer UI. For deeper analysis, query the system.ai_gateway.usage table to analyze usage metrics like token counts, latency, requester, and request tags for every request routed through the gateway (see the usage table schema).

Note

Unity AI Gateway endpoints log to system.ai_gateway.usage, whereas model serving endpoints (legacy) log to system.serving.endpoint_usage.

Common uses:

  • Monitor and optimize all your AI services from one place, across your AI stack
  • Track coding agent adoption and model popularity across teams
  • Identify the most active users, services, and models to spot optimization opportunities

See Model usage for Unity AI Gateway services.

Inspect requests and responses

Enable inference tables on a model service to log request and response payloads to a Unity Catalog table. Use inference tables to:

  • Troubleshoot failed or unexpected responses
  • Verify that guardrails are working, such as PII redaction
  • Audit sensitive-data exposure

See Monitor model services using inference tables.

Audit access

Query the system.access.audit table to prove who accessed which AI service and when. To distinguish a human caller from an agent acting on their behalf, use the run_by and run_as fields in identity_metadata.

Common uses:

  • Produce an access trail for compliance and regulatory reporting
  • Distinguish a human caller from an agent acting on their behalf
  • Maintain a central audit trail across the coding agents your teams use

See Audit log system table reference.

Track dependencies

Lineage tracks dependencies between foundation models and Model Services, as well as their downstream consumers. Use lineage to scope the impact of deprecating or changing a Model / MCP Service and notify its owners.

See Lineage in Unity Catalog.

Additional resources