Getting started

This walks through the shortest path from “I have a warehouse” to “I ran a query.”

datahashi is split into two planes, at two hosts:

• Control plane (api.datahashi.com) — your workspace, connections, models, and API keys. This is what the console talks to.
• Data (query) plane (engine.datahashi.com) — where queries actually run, over REST or MCP.

The same credential — a bearer token carrying workspace and mode — crosses from control plane to data plane on every request.

1. Create a workspace

Sign in at the console and create a workspace (called an org in the API). A workspace is the isolation boundary for your sources, models, and API keys — nothing crosses between workspaces.

2. Connect a data source

Connect the warehouse your models will query — Snowflake, BigQuery, Postgres, MySQL, or DuckDB. datahashi stores the connection’s address and credential; it never widens the access that credential already grants, and row-level security and cost limits are layered on top server-side. You can connect more than one source — each semantic model binds to exactly one.

3. Write a semantic model

A model is YAML: entities (your business nouns), dimensions and measures on those entities, and metrics built from measures. See Write your first model for a full walkthrough, or the semantic model glossary for the vocabulary first.

4. Mint an API key

From your workspace, mint an API key (POST /v1/orgs/{orgId}/api-keys on the control plane). Every key is scoped to a mode:

• governed — the default. Consumers see named metrics and curated views only.
• explore — additionally exposes raw measures and dimensions across the join graph, for internal tooling and analysts.

The key is a bearer token, shown exactly once — store it somewhere safe. You can revoke or rotate it later without disrupting your other keys.

5. Query it

Send the key as a bearer token to the data plane:

curl https://engine.datahashi.com/v1/query \
  -H "Authorization: Bearer $DATAHASHI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "metrics": ["avg_order_value"],
        "dimensions": [{ "name": "region" }],
        "order": [{ "ref": "region" }],
        "limit": 100
      }'

See Query the API for the full request/response shape, or Use with an LLM agent to wire the same credential up over MCP instead.