Metabase

The API covers card (question) management, dashboard management, query execution, and user/permission management. Agents can run saved questions via the API and retrieve results as structured data, which is the most common automation use case. The API was not originally designed as a first-class integration surface, so it has some rough edges compared to API-first products, but it is functional for the common automation patterns.

Reliability depends on the deployment model. Cloud Metabase has managed infrastructure; self-hosted deployments inherit whatever operational reliability the team provides. Query execution reliability is fundamentally tied to the underlying database connection — Metabase API latency for query results reflects the underlying database query performance.

Metabase is the most widely deployed open-source business intelligence tool and has accumulated a large user base across companies that want BI capabilities without the cost of commercial alternatives. The REST API enables agents to read query results, manage dashboards, and trigger data exports programmatically — making it a viable integration target for agents that need to surface analytics data in automated workflows.

Authentication uses API keys for cloud/paid plans and session tokens for self-hosted deployments. The session token model for self-hosted Metabase is less clean for automation — tokens expire and must be refreshed — compared to long-lived API keys. Teams running self-hosted Metabase should plan for token refresh handling in agent integrations.

Documentation has improved significantly as Metabase has grown. The API reference is complete though sometimes terse. Teams integrating agents with Metabase will find the API reference sufficient for the common patterns, with community resources filling gaps for edge cases. The open-source nature means community-contributed examples are widely available.