We'll be with you in a second

“Stop (&go)” rules

Home

RedTrack MCP Server

Contents
Introduction
What you can do with RedTrack and AI
What is MCP, and how is it different from the API?
Requirements
Connect your assistant
What it can and can’t touch

Introduction

Ask your tracker a question → get an answer. That is the whole idea behind the RedTrack MCP Server. Connect an AI assistant like Claude or ChatGPT to your account and run the platform in plain language. Pull a report, spin up a campaign, kill the losers, all without clicking through dashboards or writing a line of API code.

  • Non-technical? You will live in the first sections and never touch code.
  • Technical? The connection details, config, and limits are all here too — no separate developer doc to hunt down.

How it works

STEP 1

You ask

Type a request in plain English — no query builder, no code.

STEP 2

It picks the tools

Your assistant maps the request to the right RedTrack actions.

STEP 3

RedTrack runs it

The actions run on your account, with the same permissions you have.

STEP 4

You get an answer

A plain-language result you can read, share, or act on right away.

What you can do with RedTrack and AI

Think of the reports you build every Monday and the changes you make by hand all week. Most of that becomes a sentence. A few to try on day one:

  • “Show my top campaigns by profit this week — flag anything losing money.”
  • “Pause every campaign that spent over $500 with zero conversions.”
  • “Which countries have the best ROI on my nutra offer?”
  • “Compare Meta and TikTok over the last 30 days.”
  • “How did my Shopify source perform last month?”

The server exposes more than 40 actions across every major part of RedTrack. Here is the range, with a real question for each area:

Ask about…
A question you can type
Type
Reporting
“Show my top campaigns by profit this week.”
Read
Campaigns
“Pause every campaign that spent over $500 with zero conversions.”
Read + Write
Traffic sources
“Create a TikTok source from the preset and clone my Meta setup.”
Read + Write
Offers
“Which offers are running on my highest-traffic campaign?”
Read + Write
Affiliate networks
“Add a ClickBank network from the preset.”
Read + Write
Landing pages
“List every prelander that hasn’t had a conversion in 30 days.”
Read + Write
Tracking domains
“Regenerate the SSL on my tracking domain.”
Read + Write
Streams & optimization
“Show optimizer winners for stream X and reset the losers.”
Read + Write
Clicks & conversions
“Pull the raw log for this click ID so I can debug attribution.”
Read
Cost sync
“Why aren’t my Meta costs updating?”
Read
Every action is either a Read (look only) or a Write (create, update, delete). Your assistant confirms with you before anything destructive — more on that below.

What is MCP, and how is it different from the API?

MCP (Model Context Protocol) is an open standard that connects AI assistants to outside tools and data. Anthropic introduced it, and it now works across Claude, ChatGPT, Cursor, Codex, and more. Think of it as a universal adapter between your assistant and RedTrack. New to it? The protocol’s own getting-started guide is the clearest starting point.

RedTrack has always had a REST API for developers. The MCP Server does not replace it, it runs on top of the same API and serves a different job.

Aspect
Direct REST API
MCP Server
Who it’s for
Developers
Anyone with an AI assistant — no code
How you use it
Write and maintain code that calls endpoints
Ask in plain language; the assistant calls the right tools
Multi-step tasks
You orchestrate every call
The assistant chains steps (find losers, then pause them)
Setup
API key plus custom integration code
API key plus one URL in your AI client
Best for
Custom apps, dashboards, pipelines
Analysis, quick changes, exploration
Same engine, different steering wheel.
The API is for building software. MCP is for getting things done with an assistant. Both run on the same data and respect the same account permissions.

Requirements

Three things, and you are ready to connect:

You need
Details
A RedTrack account with API access
API/MCP access must be enabled on the plan. Check this in your Profile settings → Subscription.
An MCP-capable AI client
Claude Desktop or Claude Code, ChatGPT, Cursor, Codex, Antigravity — anything that speaks MCP over HTTP or SSE.
Your RedTrack API key
Generated in RedTrack → Settings → API. Treat it like a password.

Connect your assistant

There are two ways to set it up:

  • The easy way pastes one URL into your assistant, no files.
  • The technical way uses a config file with your key kept in an environment variable. Both take about five minutes.

OPTION 1 · EASIEST — NO CODE

Paste a URL into your assistant

Best if you just want it working. Uses the API key from Requirements above.

  1. Put your key on the end of the server URL: https://mcp.redtrack.io/mcp?api_key=<your-api-key>
  2. In your AI client (Claude Desktop, ChatGPT, Cursor, and others), open the connector or MCP settings and add a custom connection.
  3. Paste the URL → save → start a new chat.

! Keep it private: paste the key into your client’s own settings, never into a shared chat or a file you commit.
! Test it: ask “Show my top 5 campaigns by revenue this week.” A ranked list back means you’re in.

OPTION 2 · CLAUDE CODE — CONFIG FILE

Add it with a config file, key in an environment variable

Best if you keep tools in config files. Your key stays in an environment variable, so it never lands in the file or in version control.

1. Put your key in an environment variable so it persists. Add this to your shell profile (~/.zshrc or ~/.bashrc), then restart your shell:

export REDTRACK_API_KEY="your-api-key"

2. Add the server to your project’s .mcp.json. Claude Code expands ${REDTRACK_API_KEY} at runtime, so the real key never enters the file:

{
  "mcpServers": {
    "redtrack": {
      "type": "http",
      "url": "https://mcp.redtrack.io/mcp?api_key=${REDTRACK_API_KEY}"
    }
  }
}

3. Save, then run /mcp in Claude Code to confirm the connection and list the tools.

Prefer the command line? claude mcp add --transport http redtrack "https://mcp.redtrack.io/mcp?api_key=$REDTRACK_API_KEY" adds it from your shell. Other config-file clients (Cursor, Codex, and similar) support the same environment-variable idea; check their MCP docs for the exact file.

What it can and can’t touch

The assistant only reaches the account tied to your API key, and it respects that account’s permissions. Read operations run freely. Create, update, and delete operations change your account — so your assistant confirms before anything destructive.

Deletes are permanent.
Delete actions for sources, landings, networks, domains, and streams remove data for good. Read the confirmation before you approve it.

A few limits worth knowing. Analytics reporting is rate-limited to 20 requests per minute, which is fine for interactive use, but pace heavy automation. Keep report queries to roughly 90 days per call for speed. And the server manages the RedTrack side only, it doesn’t drive the ad platforms’ own ad managers.

Connect it once, then replace this Monday’s report pull with a single question. That’s the fastest way to see what it does.