MCP server
Attribution MCP server
The Attribution MCP server connects Claude, ChatGPT, Cursor and any MCP-compatible assistant to your Attribution data, so you can ask about profit, ROAS, CAC payback and conversion paths in plain English and get the same numbers as your dashboard. It is read-only, signs in with OAuth, and is included on every plan, trials included.
Ask in Claude. Use the dashboard for setup.
Server URL
https://mcp.attributionapp.com/mcpSet up in four steps
- Step 1
Set up tracking
The assistant answers from your Attribution data, so it needs data to answer with. Start a 14-day free trial, install the Attribution tracking snippet or connect Shopify or Twilio Segment, and connect your ad accounts. If you are already an Attribution customer, skip this step and connect your assistant.
- Step 2
Add the server to your assistant
In Claude, open Settings → Connectors → Add custom connector and paste the server URL. In ChatGPT, open Settings → Connectors → Add a connector and choose MCP Server. In Cursor, open Settings → MCP → Add new MCP server and choose HTTP. Most assistants will also add it for you if you ask them to.
- Step 3
Sign in and approve read-only access
Your assistant opens Attribution's sign-in page. Choose the Attribution project the assistant should see, review the read-only scopes it is asking for, and click Approve. There are no API keys to copy and nothing to install.
- Step 4
Ask your first question
Start with "Connect to my Attribution project and tell me what you can see." The assistant learns your saved views, channels and events, then answers in your own names. From there, ask what you would ask an analyst.
Instructions for Claude Desktop, Claude Code, ChatGPT, Cursor and other clients are in the MCP server docs.
What you can ask
The assistant answers from the same reports as your dashboard. Here is what it can do today, with questions you can paste in as they are. The full reference is in the MCP tools docs.
Your marketing P&L by channel
Spend, cost per conversion, ROAS, conversions, revenue and profit for every channel and sub-channel, paid and unpaid, in one table with a total row. The same numbers as your dashboard.
- “Get last month's dashboard report. Which channels are profitable, and where are we losing money?”
- “Which three channels should we scale, and which should we cut, based on profit?”
What changed between two periods
Two periods side by side with the differences worked out for you: the biggest profit gains, the biggest drops, and any channel whose profit crossed zero.
- “Compare this quarter to last quarter. Which channels gained the most profit and which dropped?”
- “How does this month compare to last month, and what moved the most?”
Cohorts and CAC payback
Revenue by acquisition cohort over time, with the age at which each cohort paid back its spend, so you can see whether recent cohorts are getting better or worse.
- “Show me monthly acquisition cohorts for the last year. How many months does a cohort take to pay back its spend?”
- “Are our recent cohorts paying back faster or slower than older ones?”
New customers versus repeat buyers
Splits revenue into first-time and repeat purchasers by channel, with the cost of acquiring a new customer on each, so you know which channels grow the base.
- “For last quarter, which channels brought in new customers rather than repeat buyers?”
- “What did a new customer cost on each paid channel last month?”
Conversion paths
The touchpoint sequences that lead to conversions, most frequent first, with revenue and time to convert, so you can see which channels assist and which close.
- “What are the most common paths to purchase this quarter?”
- “How many touchpoints does a typical customer go through, and which channels assist rather than close?”
Your project, channels, views and events
The assistant starts by learning your project: your saved views, attribution models, channel and campaign tree, and the events you track, so it uses your names instead of guessing.
- “Connect to my Attribution project and tell me what you can see.”
- “Which events do we track, and which of them carry revenue?”
The connection itself
Check which project the assistant is connected to, end the session when you are done, and flag a report that looks wrong so Attribution can check it.
- “Which Attribution project are you connected to?”
- “That report looks wrong to me. Log it for Attribution to check.”
Why the answers hold up
An assistant is only as good as the data it reads. Attribution binds ad spend to the individual people who clicked, deduplicates conversions across every channel, and ties revenue back to the touchpoints that produced it. So when the assistant says a channel lost money last month, that figure comes from user-level cost data, not from the conversion counts each ad platform reports about itself.
The answers are also easy to check. The report tools return results that are already calculated, so the assistant quotes numbers instead of doing arithmetic, and most reports include a link to the same report on your dashboard. Every number traces back to the visits, costs and revenue events underneath it.
Run MMM and incrementality analysis in your assistant
Because the assistant can pull spend, conversions and revenue by channel and by period, you can use it to run media mix modeling and incrementality analysis on your Attribution data, on every plan, without exporting anything. Ask it to look at how revenue responds as spend on a channel rises, or to compare results before and after a change. Read more about media mix modeling and incrementality testing with Attribution.
Security and privacy
- Read-only. The assistant can run reports and read your setup. It cannot change settings, filters, reports or integrations.
- OAuth sign-in. You sign in to Attribution and approve access yourself. There are no API keys to paste into a chat.
- Scoped to your project. You choose which Attribution project the assistant connects to when you approve it.
- Revocable at any time. Revoke a session in Account → Security in your dashboard, or ask the assistant to disconnect.
- Your data and the assistant's provider. Report results go to the company that runs the assistant and follow its terms. Attribution does not use your data to train AI models.
Frequently asked questions
The Attribution MCP server, answered.
The assistant answers from your Attribution data, so it needs data to answer with. Start a 14-day free trial, install the Attribution tracking snippet or connect Shopify or Twilio Segment, and connect your ad accounts. If you are already an Attribution customer, skip this step and connect your assistant. Then add the server URL, https://mcp.attributionapp.com/mcp, in Claude, ChatGPT, Cursor or another MCP-compatible assistant and sign in with OAuth.
The Attribution MCP server is a connection between your Attribution account and an AI assistant, built on the Model Context Protocol (MCP), an open standard for connecting assistants to data. Once connected, your assistant can run the same reports as your Attribution dashboard and answer questions about spend, revenue, profit, ROAS, CAC payback and conversion paths in plain English.
Claude, ChatGPT, Cursor and any MCP-compatible assistant. You add the server URL, https://mcp.attributionapp.com/mcp, sign in with OAuth, and start asking. Step-by-step instructions for each client are in the MCP server docs.
Yes. The Attribution MCP server is included on every plan and switched on for every account, trials included. There is no separate fee and no add-on to buy. Every self-serve plan starts with a 14-day free trial.
No. Every data tool is read-only. The assistant cannot change settings, edit filters, modify reports or alter integrations. The only action it can take is ending its own session when you ask it to disconnect.
Yes. The report tools run the same computation as the dashboard pages and read their settings from your saved views, so the assistant reports the same spend, conversions, revenue and profit you see in Attribution. The figures come back already calculated, and most reports include a link to open the same report on your dashboard so you can check them.
When the assistant runs a report, the result is sent to the company that operates the assistant, such as Anthropic for Claude or OpenAI for ChatGPT, and from then on it is governed by that provider's terms. Attribution does not use your data to train AI models. Avoid connecting projects that hold regulated data, such as health records.
Ask the assistant to disconnect from Attribution, or open your Attribution dashboard, go to Account → Security, find the assistant under Active Sessions and click Revoke. Its access stops immediately.
Yes, for setup. Ask in Claude. Use the dashboard for setup. You connect ad platforms, define conversion events, organize channels and save views in the dashboard. Day to day, you can ask the assistant instead of building reports by hand.
Ask your marketing data anything
Start a 14-day free trial and connect your assistant on day one, or book a demo and we will walk you through it.
