> ## Documentation Index
> Fetch the complete documentation index at: https://docs.metricanic.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect an AI agent

> Give Claude read-only access to campaign setup and reports, with enough context to interpret the numbers correctly.

Connect an AI agent when you want to investigate campaign performance in conversation. Instead of exporting rows and explaining every metric yourself, the agent can read Metricanic's report fields, campaign configuration, goals, and aggregate results through MCP.

The connection is read-only. The agent can explain findings and suggest changes, but you apply those changes in the tracker. It cannot edit campaigns, change rotators, create automation rules, or pause advertising.

## Choose the connection you need

| Goal                                           | Connection                                                      |
| ---------------------------------------------- | --------------------------------------------------------------- |
| Explain how Metricanic works                   | [Open a guide in ChatGPT or Claude](/index#ask-about-a-guide)   |
| Analyze your workspace's campaigns and reports | Connect the Metricanic product MCP below                        |
| Run your own scheduled export or application   | Use the [REST API](/api-reference/introduction) with an API key |

Documentation access does not grant access to your workspace. The product MCP requires a separate authorization and reads data for the workspace you connect.

**Claude** is available in **AI Agents**. **ChatGPT** is shown as coming soon. Product MCP access requires an eligible Pro or Scale subscription.

## Connect Claude

1. Open **AI Agents** in Metricanic and click **Connect AI Agent**, then choose **Claude**.
2. In Claude, open **Customize → Connectors**.
3. Click **+**, then **Add custom connector**.
4. Enter `Metricanic` as the name and the following remote MCP URL:

```text theme={null}
https://mcp.metricanic.com/mcp
```

5. Add the connector and click **Connect**. Sign in to Metricanic if prompted, then allow the connection.
6. Return to **AI Agents** and check that the connection appears.

Authorization creates a dedicated read-only connection. You do not paste a public API key into this flow.

## Start with a simple request

Use your campaign's name and say what you want to achieve:

> Analyze my Lead Search campaign and help me make it profitable.

Replace **Lead Search** with your campaign's name. Add a date range if you want to review a specific period. You do not need to list tools, metrics, or report dimensions in the first message.

A useful answer should explain where the campaign loses money, suggest what to try first, and show the numbers behind that suggestion. Check how much data supports each recommendation. One visit and one conversion are not enough to establish a reliable trend. The agent suggests changes, and you apply them in the tracker.

<Accordion title="What the agent can read">
  The agent has tools to discover entities, read campaign and rotator setup, inspect goal definitions, query aggregate reports, and summarize event health. It can also read the supported automation options. That automation catalog describes what rules can do. It does not expose your saved rules or run them.

  For technical prompts, start with `capabilities` and `get_metricanic_context`. MCP report inputs name source parameters `trafficParam1` through `trafficParam15`. The REST API uses `token1` through `token15` for the corresponding report dimensions.
</Accordion>

## Check the answer against the tracker

Open the same campaign report with the same dates and timezone. Compare the totals, then the specific segment the agent identified. If revenue is missing, inspect the goal's revenue settings and event health before treating a negative ROI as a traffic decision.

If the agent cannot find the campaign, confirm which workspace you authorized and ask it to list available campaigns. If access stops after a plan change, check **Subscription** as well as the connection status.

To remove access, use **Archive AI agent** in **AI Agents**. This revokes that connection. Google integrations and public API keys are managed separately.