Skip to content

Data boundaries

This page explains what files dbt-tools reads, what data may appear in its output, and what it does not access.

What dbt-tools reads

dbt-tools reads dbt artifact files from the configured target root. It does not connect to your warehouse, dbt Cloud, or any external service unless you explicitly provide a remote target root.

FileRead by dbt-toolsPurpose
manifest.jsonYesModel definitions, test definitions, dependencies, metadata
run_results.jsonYesExecution status, timing, error messages
catalog.jsonYes (optional)Column-level type and description metadata
sources.jsonYes (optional)Source freshness results
semantic_manifest.jsonNoNot currently used
Warehouse credentialsNoNever read by dbt-tools
dbt profiles (~/.dbt/profiles.yml)NoNever read by dbt-tools
dbt Cloud APINoNot accessed

What may appear in output

CLI JSON output, Web UI views, and MCP tool responses may include any data that is present in the artifact files. This includes:

Data typeSourceExample
Model names and pathsmanifest.jsonmodel.my_project.fct_orders, models/core/fct_orders.sql
Test namesmanifest.jsontest.my_project.not_null_orders_id
Column names and typesmanifest.json, catalog.jsonorder_id, VARCHAR(256)
Model descriptionsmanifest.jsonFree-text descriptions from schema.yml
Execution error messagesrun_results.jsonSQL errors, schema errors, assertion failures
Execution timingrun_results.jsonStart time, end time, duration
dbt version and schema versionBoth1.8.0, v11
Project name and invocation IDmanifest.jsonmy_project, UUID
Adapter typemanifest.jsonbigquery, snowflake
Generated timestampsBothISO 8601 timestamps
Environment metadatamanifest.jsonAny DBT_ENV_CUSTOM_ENV_* variables set at dbt run time

Environment metadata

dbt captures environment variables prefixed with DBT_ENV_CUSTOM_ENV_ and stores them in manifest.json under the metadata.env field. These values appear in dbt-tools output.

Do not set credentials, tokens, or secrets as DBT_ENV_CUSTOM_ENV_* variables. If a secret is set this way, it will appear in manifest.json and in all dbt-tools output that reads the manifest.

What dbt-tools does not access

  • Your data warehouse (no SQL queries are issued)
  • dbt Cloud (no API calls)
  • Your dbt profiles or connections (~/.dbt/profiles.yml)
  • Any files outside the configured target root directory or prefix
  • The internet (unless --dbt-target points at a remote S3 or GCS prefix, which requires explicit configuration)

Web UI data handling

The Web UI (@dbt-tools/web) is a server that runs locally. It reads artifacts from the configured target root and serves the browser UI from the same machine. The browser communicates only with the local server; it does not make external requests to third-party services.

If you deploy the Web UI to a remote server, the artifact data is accessible to anyone who can reach that server. Use appropriate network controls (firewall rules, authentication proxies) for remote deployments.

MCP tool responses

The MCP server returns data from artifact files to the connected AI client. The client may include that data in its context window, summarize it, log it, or transmit it to third-party services depending on the client's own configuration and privacy settings.

Review the privacy policy and data handling documentation of your AI client before connecting it to production artifacts.

Released under the repository license terms.