MCP Server

AI marketing attribution software for Claude, ChatGPT, and Cursor

SourceLoop's MCP server connects your lead source, conversions, and revenue data to any AI agent. Ask about attribution in plain English, right inside Claude, ChatGPT, or Cursor. No SQL, no exports, no code.

Get started free Book a demo

Free with any SourceLoop plan.

Which channels drove the most leads this quarter, and how do they compare to last quarter?

Calling sourceloop.leads_by_channel 0.4s

Q1 brought in 1,842 new leads, led by Google Ads at 612. LinkedIn Ads improved the most quarter over quarter, up 38%, on the back of three retargeting campaigns. Here is the breakdown:

Leads by channel · Q1 2026 SourceLoop · last_touch

Google Ads612

Meta Ads438

LinkedIn Ads357

TikTok Ads271

Organic search164

Want me to open these channels in your dashboard or push the top performers to Google Ads as offline conversions?

Where did our top 5 customers come from?

SourceLoop Claude Sonnet 4.6

Used by 750+ marketing teams, agencies, and SaaS companies

SourceLoop MCP use cases

Ask anything about your attribution. Get a real answer.

SourceLoop's MCP server exposes the same live data your dashboard runs on. Every chart and every number an agent returns is traceable back to source, never a guess.

Reporting

Build any report in one message, no SQL, no exports.

Content

See which content drives pipeline, ranked by revenue, not traffic.

Conversion sync

Push closed deals to your ad platforms, matched automatically.

CRM

Keep your CRM attribution clean, synced to HubSpot, Salesforce, Pipedrive.

Metrics

Track any metric over time, conversion rate, CAC, or pipeline.

Compatible clients

Works with the AI tools you already use

SourceLoop's MCP server speaks the open standard, so any agent or LLM that supports MCP can query your attribution data, no custom integration required.

Claude
ChatGPT
Cursor
Windsurf
Zed
Perplexity

And any MCP-compatible agent via JSON-RPC over stdio or SSE.

FAQ

MCP server, frequently asked

What is MCP, and why does SourceLoop support it?

Model Context Protocol (MCP) is an open standard, originally developed by Anthropic, for connecting AI agents to live tools and data. Any MCP-compatible agent (Claude, ChatGPT, Cursor, Windsurf, and more) can call into SourceLoop the same way they call any other MCP server. We support it because the moment your team starts using AI agents to think about marketing performance, the agents need real attribution data, not stale screenshots or exports.

Which AI tools does this work with?

Claude Desktop, ChatGPT, Cursor, Windsurf, Zed, Perplexity, and any LangChain or LlamaIndex agent. Custom JSON-RPC clients also work, if it speaks MCP over stdio or SSE, it can query SourceLoop.

What can agents actually query?

The same metrics the dashboard exposes: source, medium, campaign, ad ID, landing page, full conversion path, channel performance, pipeline, revenue, customer-level attribution, and offline-conversion sync status. Same data layer, same filters, no separate API surface to learn.

Is this safe? Can the agent leak data?

Auth is via scoped API keys, one per agent or workspace. You can revoke any key in one click. Every call is logged with the prompt, the resolved query, and the response, so you have a full audit trail. No third-party tokens, no shared credentials.

Can the agent write back to my SourceLoop, my ad platforms, or my CRM?

Yes, with explicit write scopes. Tag a lead, push an offline conversion to Google Ads or Meta, flag a campaign for review, or sync a contact to HubSpot or Salesforce. All write actions are recorded in the audit log and can be disabled per key.

How do I set it up?

Generate an MCP API key from your SourceLoop settings, drop the snippet into your Claude or Cursor MCP config, restart the client. Most teams are live in under five minutes. No SDK install, no scaffolded code.