Executive brief
dbt-mcp is a tool used to interact with dbt (data build tool) via the Model Context Protocol. A vulnerability was identified where sensitive information, including raw SQL queries and database credentials, was automatically sent to the vendor's telemetry service without being hidden or removed. This could result in the exposure of private database structures or login information to the service provider if telemetry was not manually disabled.
Technical details
The vulnerability exists in the `DefaultUsageTracker.emit_tool_called_event()` function within `src/dbt_mcp/tracking/tracking.py`. The component serializes the complete arguments dictionary of every MCP tool call and transmits it to the dbt Labs telemetry service via `dbtlabs_vortex.producer.log_proto` without redaction. This includes sensitive fields such as `sql_query` from the 'show' tool and `vars` (which often contain credentials) from 'run', 'build', and 'test' tools. Telemetry is enabled by default unless specifically opted out via environment variables. An attacker with low privileges who can trigger tool calls could cause the leakage of sensitive configuration or query data to the telemetry endpoint. The issue is fixed in version 1.17.1 by implementing argument redaction.
Affected products
- dbt-labs dbt-mcp < 1.17.1
Timeline
- 2026-05-05: patched: Fix committed and version 1.17.1 released.
- 2026-05-13: advisory: GitHub Security Advisory published.
- 2026-07-16: disclosed: CVE published to NVD.