dbt-toolsDebug, inspect, and automate dbt runs from artifacts

Turn dbt manifests, run results, and metadata into deterministic insights for operators, automation, and agents.

Debug failed runs

Identify failed models and tests, trace error context, and find the downstream blast radius—from the command line or a browser.

One artifact contract, four surfaces. Use dbt-tools from shell scripts, long-lived agent sessions, a browser UI, or AI coding assistants.

Read manifest.json and run_results.json from a local target directory, an S3 prefix, or a GCS prefix with the same commands.

Debug a failed run — find failed nodes, error messages, and downstream impact
Investigate slow models — rank execution bottlenecks and inspect timing metadata
Find model impact — trace upstream and downstream dependencies before a change
Generate a CI health summary — produce deterministic JSON in GitHub Actions or CI
Ask an AI agent about a dbt run — use MCP or agent skills safely

Interface	Use when
CLI (`@dbt-tools/cli`)	Shell, CI, scripts, JSON output
Web (`@dbt-tools/web`)	Browser-based lineage and run investigation
MCP (`@dbt-tools/mcp`)	Long-lived AI agent sessions
Agent skills	Named operations for Cursor, Codex, Claude Code

dbt-tools reads manifest.json and run_results.json from:

Before using production artifacts with agents or in shared environments: Trust & Safety →