01
Connect your AI client
Authorize a compatible ChatGPT or Claude client to use Metricanic's MCP reporting tools for the workspace you enable.
AI-powered campaign analysis for media buyers
An AI ad tracker that lets ChatGPT or Claude analyze your campaigns
Connect a compatible AI client directly to Metricanic. Ask for a profitability diagnosis or full campaign report, let the model pull the breakdowns it needs, and get an evidence-based optimization plan in the same conversation.
AI Agents are included on Pro and Scale. The connection is analysis-only.
Campaign analyst
Connected to Metricanic reporting
Analyze “PopAds - US - Antivirus” and tell me whether the campaign can become profitable. Show the evidence and the next tests in priority order.
Campaign overview
Placement breakdown
Funnel comparison
Optimization plan
What “AI ad tracker” means here
Your tracker stays the source of truth. The model becomes the analyst.
Metricanic is a full paid-traffic tracker first: campaign links, routing, landers, offers, events, goals, postbacks, cost, revenue, profit, and 100+ reporting dimensions. Its AI layer gives the model controlled access to reporting tools, so its answers can be grounded in the campaign data instead of a pasted CSV or a generic marketing prompt.
02
Ask the campaign question
Name the campaign and outcome you care about: profitability, a sudden CPA change, weak placements, scaling opportunities, or tracking health.
03
The model investigates
The model chooses which reports, time ranges, dimensions, filters, and comparisons it needs instead of following a single canned checklist.
04
Review the recommendation
Get a campaign report, the evidence behind each conclusion, and a prioritized optimization plan. You decide what to apply.
Open connection standard
Metricanic includes an MCP server
Model Context Protocol (MCP) is the connection layer between Metricanic and a compatible AI client. In practical terms, it lets ChatGPT or Claude request tracker reports during a conversation instead of making you export, reformat, and paste campaign data by hand.
Metricanic gives the model instructions for using the available reports. The model can then decide which dimensions to compare and what follow-up data it needs for the question. This makes the workflow flexible without pretending every campaign can be solved by the same fixed list of checks.
Not a generic “AI summary” button
- The model can request additional report breakdowns when the first result is not enough.
- It can compare traffic segments across the dimensions media buyers use to make optimization decisions.
- It can explain when the evidence supports a cut, a scale test, more data collection, or a new lander or offer.
- You can ask follow-up questions and challenge the conclusion inside the same analysis thread.
Product demos
Watch Claude analyze a CPA campaign
These recordings show the actual workflow: connect Claude to Metricanic, ask for campaign optimization, and watch the model pull reporting context and build its recommendation.
Claude AI optimizes a CPA affiliate campaign
A narrated seven-minute walkthrough comparing the normal manual optimization process with Claude's campaign analysis through Metricanic.
Connect Claude to Metricanic and run an analysis
A short setup and output demo showing how the Claude connection is added and what the model returns when asked to optimize a campaign.
Prompts to start with
Ask for a decision, not just a dashboard recap
A useful prompt names the campaign, the business question, the evaluation window, and the level of evidence you expect. The agent can ask for more context when the available sample is not conclusive.
Prompt 1
Analyze the campaign “PopAds - US - Antivirus” and advise me how to make it profitable. Provide a full campaign report.
Prompt 2
Which publishers spent enough to evaluate but have no conversions, and which segments still need more data?
Prompt 3
Compare the last seven days with the previous seven. Explain the profit change by source, geo, device, lander, and offer.
Prompt 4
Check this campaign's conversion and postback signals. Flag anything that looks like a tracking problem rather than a traffic problem.
Designed for controlled analysis
The model advises. The buyer stays in control.
No write access
The current connection is designed for analysis and recommendations. The agent does not edit campaigns or change budgets.
Limited reporting context
Access is designed around customer-enabled workspace reports while excluding sensitive operational data, secrets, and raw visitor data.
Human-reviewed decisions
The media buyer can challenge the reasoning, ask for supporting breakdowns, and approve or reject every suggested move.
Metricanic controls its side of the connection. Data handling by the AI client is also governed by the provider and account settings you choose; review those terms before connecting campaign data.
AI ad tracker and MCP questions
How ChatGPT and Claude connect to Metricanic, what the model can analyze, and where the guardrails are.
What is an AI ad tracker?
An AI ad tracker combines campaign tracking data with an AI analysis workflow. In Metricanic, a compatible ChatGPT or Claude client can query reports, compare segments, explain what changed, and recommend actions while Metricanic remains the campaign data source.
How does Metricanic connect ChatGPT or Claude to campaign data?
Metricanic provides an MCP server for eligible accounts. MCP is an open connection standard that lets a compatible AI client request approved tracker reports as it works through a question in plain language.
Does the AI only run a fixed set of canned checks?
No. Metricanic describes the available reporting tools and how to use them. The connected model decides which reports and breakdowns it needs for the specific campaign question instead of being limited to a fixed sequence of browser, offer, or publisher checks.
Can the connected AI change campaigns?
The current Metricanic AI analysis connection is designed without write access. It can analyze allowed reporting context and recommend changes, but the media buyer reviews the evidence and decides what to apply.
Does Metricanic send secrets or raw visitor data to the model?
The connection is designed to limit access to customer-enabled workspace data and exclude sensitive operational data, secrets, raw visitor data, and write access. The AI provider's own data terms still apply to the client you choose.
Which plans include AI Agents?
AI Agents are included in the current Pro and Scale plans. Check the pricing page for current limits and plan details.