Metroflow Docs
Get productive in under 10 minutes
Self-host the agentic control plane, connect your warehouse and orchestration tools, and open a live semantic graph in the workspace. No row movement. Metadata only.
Start here
Pick a guide based on what you need right now. Each path assumes a local Docker install and points to deeper references when you're ready.
Quickstart
Clone the repo, configure .env, and bring up the workspace with Docker Compose in under ten minutes.
Connectors
Wire Snowflake, dbt, Airflow, BigQuery, Postgres, and 35+ more sources into a unified metadata graph.
Browse connectors → DevelopersREST API
Programmatic access to lineage, metrics, catalog search, connector sync, and agent queries.
API reference →Guides
Deeper product and operations guides: workspace workflows, AI agents, deployment, and authentication.
Workspace
Navigate the semantic graph, run impact analysis, and collaborate with stack-aware AI agents.
Workspace guide → AIAI & agents
Stack-aware agents that cite real assets: lineage answers, dbt suggestions, and certified metrics.
Agents guide → OperationsDeployment
Production hardening: Kubernetes, scaling the crawler, backups, and environment configuration.
Deployment guide → SecurityAuthentication
API keys, SSO, role-based access, and securing the workspace in your VPC.
Auth guide →How Metroflow fits your stack
Metroflow sits above your existing data plane, not inside it. Crawlers read metadata from warehouses, transformation projects, and orchestrators, then materialize a living graph your team and agents can reason over.
Nothing copies row data out of Snowflake or Postgres. The crawler ingests schemas, manifests, DAG definitions, and metric metadata, then keeps the graph fresh on a schedule you control.
- Sources: Snowflake, dbt, Airflow, Supabase, Dagster, and more
- Crawler: Scheduled metadata ingestion inside your network
- Graph: Cross-system lineage, ownership, and certified metrics
- Workspace: UI and agents that understand your stack end to end
What you'll build
By the end of the quickstart path, you'll have a working control plane, not a demo stub.
- A self-hosted Metroflow instance API, crawler, and workspace UI running locally or in your VPC, fully under your credentials and network policies.
- A connected semantic graph At least one warehouse and transformation source synced, with tables, models, and dependencies linked in the graph view.
- Agent-ready context Stack-aware AI that can answer lineage questions, suggest dbt tests, and cite real assets, not generic SQL guesses.