In-Depth & How-To

How Metroflow works, step by step

Connect sources, build the graph, certify metrics. Metadata only. Under five minutes to first insight.

← Documentation

Infrastructure

Understand your stack in 3 steps

Step 1 · Connect

Point at what you already run

Snowflake, dbt, Airflow, and 40+ connectors. Read-only metadata crawls. Credentials stay in your network.

OAuth or key-based auth per connector
docker compose up → first source live fast

Step 2 · Build graph

Normalize into one semantic layer

fct_revenue links to its Snowflake table, Airflow task, and Looker explores automatically. Incremental diffs keep it fast.

Manifest parsing for dbt Core and Cloud
Graph API for custom tools

Step 3 · Query & act

Lineage, impact, agents

Rename a column and preview blast radius. Ask Brain why a dashboard is stale. Daily workflow, not a wiki project.

Stack-aware editors and cited agents
RBAC per environment

Workspace features →

Metrics

Canonical metrics in 3 steps

Define

Certify on the semantic layer

Grain, filters, and stewards in one object. Lineage to source columns inherited from the graph.

Templates: MRR, churn, DAU
Preview PII exposure before certifying

Standardize

Push to every surface

BI, notebooks, APIs, and agents resolve to the same canonical ID. Override detection flags rogue SQL.

Notify consumers when definitions change
Staging metrics never leak to exec views

Reconcile

Verify and audit

Cross-check against Stripe, CRM, and ERP. Export trails for finance close and governance.

Alerts when delta exceeds threshold
Brain explains gaps with pipeline context

Metrics layer →

Go deeper

More resources

Quickstart