> AI agents: this is one page of nextanalytica's docs. The full index for agents: https://nextanalytica.app/llms.txt

# FAQ and advanced setup

> What people ask before they connect an agent: cost, clients, keys, projects, where data goes and safety, answered straight.

Updated: 2026-11-18

The questions people ask before they let an agent near their analytics, answered straight. New here? Start with [Your agent](/docs/model-context-protocol).

## Frequently asked questions

### What is the nextanalytica MCP server?

A remote MCP server at `https://mcp.nextanalytica.app/mcp`. Your agent connects to it and reads your projects' analytics, sources and team context through [its tools](/docs/model-context-protocol/tools).

### Is it really free?

Yes. Using the MCP server is free, with **30 days** of history. Your agent's own usage counts against your own plan with your provider; that part is between you and them.

### Will it work with my client?

If your client connects to a remote MCP server over streamable HTTP, yes. We check Claude Code and the Claude app ourselves. The pages for Codex, Cursor, VS Code, Windsurf and Zed follow each client's own way of adding a remote server.

### Do I need a key?

No. Most clients sign in in the browser (OAuth). An [agent key](/docs/model-context-protocol/agent-keys) is for clients and machines where a browser isn't at hand.

### I have several sites. Do I connect each one?

No. One connection covers every project you can see. Name the site in your question and your agent finds it with `list_projects`.

### Where does my data go?

From our servers to your agent, and from your agent to the company that runs it (Anthropic, OpenAI or another), under your agreement with them. We run no model on it and never use it to train one.

### Is it safe to let my agent write?

Most tools only read. The ones that write (a note, a memory, conversions, a cohort) say **Changes data**, and forgetting a visitor or changing conversions asks you to confirm first. Our tools never change your website.

> **Warning:** Text on web pages, in issues or in data can try to steer an agent (prompt injection). Read what your agent proposes before you approve a change to your code or an action through a connector.

### Can I see what agents did?

Every action through a connector is logged in **Connectors**, with who did it.

## Advanced setup

### How do I use an agent key instead of the browser?

Send the key as a header in your client's MCP settings:

```json
{ "headers": { "Authorization": "Bearer na_key_…" } }
```

Where that goes differs per client: see its page under **Clients**.

### How do I connect more than one account?

Add the server twice under two names, each signed in or with its own key: `nextanalytica` and `nextanalytica-work`.

### How do I remove the connection?

Remove the server in your client (`claude mcp remove nextanalytica` in Claude Code). To cut it off from our side, revoke the key in **Access**, or remove the person from the project.

## Keep exploring

- [Your agent](/docs/model-context-protocol)
- [Team access](/docs/model-context-protocol/team-access)

Canonical address: https://nextanalytica.app/docs/model-context-protocol/faq
