Coupler.io MCP Server

Coupler.io MCP Server is the official Model Context Protocol integration for the Coupler.io data automation platform. It bridges Coupler.io data flows (pipelines that pull data from sources like HubSpot, Salesforce, Google Ads, Facebook Ads, Google Analytics, Shopify, BigQuery, QuickBooks, Xero, and more) to AI assistants so they can answer questions about marketing, sales, finance, and ecommerce metrics without manual SQL or BI dashboards.

The server exposes four tools that let an AI agent discover available data flows, inspect their schemas, fetch execution metadata, and run read-only SQL queries against the resulting SQLite snapshots. This means an LLM can plan a question, look up which data flow contains the relevant data, examine the table structure, and query it directly, returning insights, calculations, or comparisons in plain English.

The MCP server is distributed as a Docker image (ghcr.io/railsware/coupler-io-mcp-server) and is maintained by Railsware, the company behind Coupler.io. Using it requires an active Coupler.io subscription on the Professional plan or higher, configured data flows with AI destinations, and a Personal Access Token.

Tools

Prerequisites

• A Coupler.io account on the Professional plan or higher
• At least one data flow configured in Coupler.io with an AI destination
• A Coupler.io Personal Access Token, generated at https://app.coupler.io/app/mcp/
• Docker installed and running locally

Claude Desktop configuration

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "coupler": {
      "command": "docker",
      "args": [
        "run",
        "--pull=always",
        "-e",
        "COUPLER_ACCESS_TOKEN",
        "--rm",
        "-i",
        "ghcr.io/railsware/coupler-io-mcp-server"
      ],
      "env": {
        "COUPLER_ACCESS_TOKEN": "<your_token>"
      }
    }
  }
}

Replace <your_token> with the Personal Access Token from the Coupler.io app. The --pull=always flag ensures Docker fetches the latest image on each launch. Remove it if you want to run offline against a cached image.

The same Docker command works for other MCP-compatible clients such as Cursor and VS Code by adapting their respective configuration formats.

• Ask an AI assistant for marketing KPIs (CAC, ROAS, blended CPA) across Google Ads, Facebook Ads, and Google Analytics data flows without opening a BI tool
• Generate sales pipeline reports from HubSpot, Pipedrive, or Salesforce data flows and have the LLM summarize movement by stage or rep
• Run finance queries on QuickBooks or Xero data flows, such as month-over-month profit comparisons or aged accounts receivable
• Analyze ecommerce performance from Shopify data flows, comparing product revenue, refund rates, or cohort behavior
• Combine cross-source metrics in a single conversation, for example matching ad spend from Facebook against attributed revenue from Shopify

• "List all my Coupler.io data flows and tell me which ones have ad spend data."
• "From my Google Ads data flow, show me the top 10 campaigns by ROAS for the last 30 days."
• "Compare this month's net profit to last month using my QuickBooks data flow."
• "How many open opportunities are in each stage in my HubSpot pipeline data flow?"
• "Pull the schema of my Shopify data flow, then calculate refund rate by product category."

• Officially maintained by Railsware, the team behind Coupler.io
• Unlocks 70+ (Coupler.io advertises 400+ via the platform) source integrations through a single MCP server instead of one MCP per tool
• Read-only SQL access keeps data safe while allowing flexible analytical queries
• Distributed as a Docker image, so no Node or Python toolchain is required

• Requires a paid Coupler.io subscription on the Professional plan or higher
• Only works against data flows you have already configured in Coupler.io, no ad-hoc source connections from the MCP layer
• Read-only by design, so write-back to source systems (creating HubSpot contacts, updating Shopify orders) is not supported
• Requires Docker, which adds a dependency for non-developer users

• Supermetrics MCP for marketing data extraction into spreadsheets and BI
• Fivetran and Airbyte for self-hosted ELT pipelines that can be paired with a warehouse-level MCP
• Direct per-source MCP servers (HubSpot MCP, Google Ads MCP, etc.) when you only need a single source rather than unified data flows