Skip to main content
toorow is a self-hostable marketing data platform built for AI agents. It speaks the Model Context Protocol (MCP), so an agent reads governed indicators without ever seeing a credential, a provider API, or a database table.

​
High-Level Architecture

​
Key Capabilities

  • 39 Pre-Built Connectors: Metrics from GA4, Search Console, Meta Ads, TikTok Ads, LinkedIn Ads, Shopify, Stripe, and Klaviyo converge into a canonical semantic model built with dbt.
  • Audited & Traceable Analysis: Freshness, data provenance, quality checks (DQ baselines), and alerts accompany every metric served to AI agents.
  • Isolated Multi-Tenant Access: Inbound identity, provider OAuth scopes, and dataset access are strictly isolated per project.
  • Dual-Channel Response Model: Agents receive a short, contextual text summary while the accompanying MCP widget receives a rich, structured data payload (structuredContent) for rendering visual cards.
  • Auto-Discovered Connector Architecture: Connectors are self-contained and discovered at boot, each carrying a manifest, staging SQL, dbt marts, and golden-pull conformance tests.

​
Strategic Direction

toorow is designed as a sovereign control plane for agentic analytics rather than a basic connector catalog:
  1. Managed or external warehouse ingestion.
  2. Semantic governance and MDM dictionary enforcement.
  3. Mandatory data quality and lineage tracking.
  4. Shared business context and progressive disclosure.
  5. Post-click attribution without proprietary tracking pixels.
  6. Continuous evaluation loop for agent reliability.

​
Explore the Documentation

Supported Connectors

Browse all 39 pre-built connectors available in toorow.

MCP Host Integration

Connect any MCP host over streamable HTTP.

FastMCP Agent Tools

Inspect FastMCP tool definitions and parameter schemas.

Quickstart Guide

Install and run toorow locally with uv and pnpm.