Skip to main content
toorow is a stack-agnostic MCP platform for unified marketing reporting. It implements the Model Context Protocol (MCP) so assistants like Claude can query your performance data — GA4, Search Console, Meta Ads, Shopify and more — in natural language, without burning tokens on raw JSON and without leaking one tenant’s credentials into another’s context.

​
Two ways to run toorow

toorow Cloud

We host it for you. Connect your accounts and ask your data from your AI assistant — no infrastructure to run. Docs in this section grow as the product does.

Self-hosted

Run your own instance on your own stack, extend it with your own connectors, and deploy it wherever you like. Start with the Quickstart.

​
What it does

  • Multi-tenant identity & OAuth brokering. Each user links their own accounts. Tokens are isolated per tenant, encrypted at rest, and refreshed automatically — every tool call is bound to the human who authorized it and recorded in an audit log.
  • Warehouse + semantic marts. Raw data lands in BigQuery and is harmonized through dbt, so “cost” or “sessions” mean the same thing across every source.
  • Compact MCP responses + rich widgets. The server returns a light textual summary to the model and streams the full dataset to a self-contained React widget — keeping large payloads out of the context window.
  • Governance & quality. Connection health, sync performance and agent traffic are monitored; semantic quality monitors keep analyses on fresh, trustworthy data.