Connect your AI agent to MCP

Wire up an MCP-capable client so it can call MaxoPerf’s tools — the same actions you’d take in the console, driven by natural language.

Before you start

You need a MaxoPerf API key. See Get an API key — read that page first, it covers the full-account access an API key grants and how to use it safely.

Find your endpoint

Open Settings → Connect AI tools (/settings/connect-ai-tools) in the console. The page shows your account’s MCP endpoint URL and ready-to-copy config snippets for each supported client. Your endpoint is:

https://app.maxoperf.com/mcp

If the page shows “MCP endpoint not configured for this environment,” ask an administrator to configure the MCP server origin for your deployment.

Configure your client

Claude Code — run from your terminal:

claude mcp add --transport http maxoperf https://app.maxoperf.com/mcp \
  --header "Authorization: Bearer mpak_your_api_key_here"

Cursor — add to .cursor/mcp.json (project or global):

{
  "mcpServers": {
    "maxoperf": {
      "url": "https://app.maxoperf.com/mcp",
      "headers": {
        "Authorization": "Bearer mpak_your_api_key_here"
      }
    }
  }
}

Claude Desktop / claude.ai — go to Settings → Connectors → Add custom connector, paste the endpoint URL, then paste the API key when prompted.
ChatGPT — go to Settings → Connectors → Create, paste the endpoint URL, and set the Authorization header to Bearer mpak_your_api_key_here.

Copy the exact snippet for your client from the console page — it’s pre-filled with your account’s real endpoint, so you only need to swap in the API key.

Multiple accounts

If your API key’s principal has access to more than one account, also set X-Account-Id: <accountId> — see MCP overview.

What the agent can do

Once connected, the agent can call MaxoPerf’s tools to create tests, start runs, poll status, and read KPIs/results — anything the API key is authorized to do. See Tools for the full catalog. Because the key grants full-account access, review your tool’s call log before approving destructive actions (delete run, cancel test, revoke key) — or connect in read-only mode first if you just want the agent to investigate.

Choosing what to test

Before you point an agent at a real workload, skim Executors — it explains when MaxoPerf runs an API-style load test (JMeter, k6, Taurus, …) versus a full browser test (Playwright, Selenium), which changes what the agent should generate.

Where to go next

MaxoPerf agents — the agent skill that teaches connected agents the full workflow, faster than manual config.
Tools — every tool, its exact arguments, and what it returns.
Secrets and environments — keep credentials the agent’s tests use out of test files.