Connecting an AI Agent via MCP

You can connect Adsafee to an AI assistant (for example, Claude Code) over the MCP protocol (Model Context Protocol). Once connected, the agent can read your analytics and manage your cabinet in natural language — strictly within your own access rights.

Ask it in plain language, for example:

• "Show me yesterday's invalid traffic by country."
• "Which of my integrations had the most fraud this week?"
• "Create a rule that blocks traffic from hosting providers."
• "Set up a weekly email report of my traffic."

What is MCP

MCP is an open standard that lets AI assistants connect to outside services. Adsafee runs an MCP server at https://dashboard.adsafee.com/mcp; you connect once through a normal browser sign-in — no tokens to copy or keys to paste.

Connect

Adsafee works with any AI client that supports MCP. The server address is the same everywhere:

https://dashboard.adsafee.com/mcp

On first use the client opens a browser: sign in to Adsafee (same login as the dashboard — email or Google) → consent screen "Allow client to act on your behalf?" → Approve. A 90-day token is issued; no repeated sign-in needed.

Claude Code (CLI)

claude mcp add --scope user --transport http adsafee https://dashboard.adsafee.com/mcp

--scope user makes the server available in all projects. Without it, claude mcp add registers the server for the current project only (the default scope is local). Check with claude mcp list.

Claude Desktop

Settings → Connectors → Add custom connector, paste the URL https://dashboard.adsafee.com/mcp (available on paid Claude plans). On first launch, sign in via the browser.

Cursor

Add the server to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project):

{
  "mcpServers": {
    "adsafee": {
      "url": "https://dashboard.adsafee.com/mcp"
    }
  }
}

Then in Settings → MCP, click sign-in next to the server to authorize.

VS Code (GitHub Copilot, Agent mode)

For all workspaces, run MCP: Open User Configuration (Command Palette) and add the server to the user mcp.json; for a single project use .vscode/mcp.json instead. Same shape either way:

{
  "servers": {
    "adsafee": {
      "type": "http",
      "url": "https://dashboard.adsafee.com/mcp"
    }
  }
}

Open Copilot Chat in Agent mode — confirm sign-in on first use.

Other clients

Any client that supports MCP servers (Windsurf, Zed, etc.) connects the same way: point it at https://dashboard.adsafee.com/mcp and sign in through the browser on first use.

TIP

Menu labels vary by client version — check your client's MCP docs. Only the server address and the browser sign-in are constant.

What the agent can do

The agent acts as you, with your role — it never gets more rights than you have in the dashboard.

Read (analytics):

• View traffic and fraud statistics — metrics, breakdowns, ready-made presets.
• List your accessible workspaces, integrations, rules, and reports.

Changes — in two steps:

• Plan — the agent first prepares a change and shows exactly what will happen. Nothing changes yet.
• Apply — the change runs only after the plan is explicitly applied.

Capability	Owner	Analyst
Read statistics and catalogs	✓	✓ (within their integrations)
Changes (rules, reports, integrations, team…)	✓	—

Multiple workspaces

If you belong to several workspaces, the agent will ask which one to act on. If you have just one, it is selected automatically.

Changes and subscription

Changes are available only to the owner (Owner) and only with an active subscription that allows writes. Irreversible actions (such as deletion) require an explicit confirmation in the plan.

Security and revoking access

• You sign in through your browser, the same as the dashboard — Adsafee never asks you to copy a token or paste a key.
• Your permissions are re-checked on every request: a role change, integration limit, or expired subscription instantly changes what the agent can do.
• Access lasts 90 days and can be revoked at any time; after that you simply sign in again.
• Disconnect on the client side: claude mcp remove adsafee.

Next steps

• Metrics Reference — what the figures the agent returns mean
• Team & Access — Owner/Analyst roles and integration scope
• Filtering Rules — what the agent can configure