Skip to content

Choose by goal

Use this page to route to the right dbt-tools interface for your job. If you already know which package you need, go directly to its guide.

Decision table

GoalBest interfaceWhyStart here
Check whether a run succeededCLIFast, scriptable, JSON outputDebug a failed run
Debug a failed model or testCLI + WebCLI summarizes; Web gives lineage contextDebug a failed run
Investigate slow modelsWeb + CLIWeb for visual exploration; CLI for automationInvestigate slow models
Understand model impactCLI + WebTrace upstream and downstream dependenciesFind model impact
Automate checks in CICLIDeterministic JSON output, reliable exit codesGenerate CI health summary
Move from CLI JSON to browser viewWebDeep-link from JSON output directly to the UIOpen CLI result in Web
Let an AI assistant query artifactsMCPLong-lived server with resident parsed cacheAsk an agent about a dbt run
Use agent skills in your IDEAgent skillsCursor, Codex, Claude Code primitive skillsInstall agent skills
Build custom TypeScript toolingCoreProgrammatic APICore reference
Read artifacts from object storageCLI / Web / MCPShared --dbt-target configS3 or GCS

Interface overview

CLI (@dbt-tools/cli)dbt-tools binary. Use for one-shot shell commands, CI pipelines, and scripts that need JSON output. Starts and exits in one invocation. Exit codes signal success or failure.

Web (@dbt-tools/web)dbt-tools-web binary. Starts a local server for browser-based investigation. Best for exploring lineage graphs, run timelines, and visual execution context.

MCP (@dbt-tools/mcp)dbt-tools-mcp binary. Runs a long-lived Model Context Protocol server. An AI client connects and makes many tool calls against one parsed artifact session. Best when an agent needs to follow up multiple questions about the same run.

Agent skills — pre-built skills for Cursor, Codex, and Claude Code. Installed alongside the CLI or MCP package. Let agents invoke stable named operations without writing raw CLI commands.

Core (@dbt-tools/core) — TypeScript library. Import directly to build custom tools, scripts, or services on top of the same artifact parsing layer that CLI, Web, and MCP use.

Prerequisites for all interfaces

  • manifest.json and run_results.json under a dbt target directory
  • Node.js 20+

New to dbt? See New to dbt? and dbt artifacts.

Remote targets (s3://, gs://) require additional environment variables. See Deploy.

Released under the repository license terms.