Skip to content
Pricing Open UserLimit
Get started

Connect an AI assistant (MCP)

Connect Claude, ChatGPT, Cursor, VS Code or any remote MCP client to UserLimit, and run load tests on your account by asking in plain words.

The connector URL

Every client connects to the same remote MCP server: https://mcp.userlimit.com/mcp

It uses Streamable HTTP with OAuth 2.1 sign-in, so there is no API key to copy. The first time a client connects, it opens the UserLimit sign-in page. Sign in with Google or e-mail, or create a free account there, then approve access. You can revoke it any time from your account.

Add it to your client

Client Where to add it
Claude (web, desktop, mobile) Settings → Connectors → Add custom connector. Paste the URL and select Connect.
Claude Code Run claude mcp add --transport http userlimit https://mcp.userlimit.com/mcp, then /mcp to sign in.
ChatGPT Settings → Apps & Connectors → Advanced, turn on developer mode, then Create a connector with the URL.
Cursor Settings → MCP → Add new MCP server, or use the one-click install on the MCP server page.
VS Code Command palette, MCP: Add Server → HTTP, paste the URL, or use the one-click install on the MCP server page.
Windsurf and other clients Add a remote (HTTP) MCP server with the URL. The client needs OAuth 2.1 with Dynamic Client Registration.

Menu names change between client versions. If you can't find the setting, search the client's settings for "MCP" or "connector".

Your first conversation

Describe the visit and the traffic in one message, for example:

Load test shop.example.com with 50 users: open the homepage, search for running shoes and add the first pair to the cart.

The assistant then:

  1. Checks your account and whether the domain is verified. If not, it adds the domain and tells you which DNS record or meta tag to publish. See Verify a domain.
  2. Writes the scenario as a short Playwright script and sends one free session through it in a real browser. If a step fails, it fixes the step before going further.
  3. Sizes the test and shows you a summary: the site, users, duration and the hours it will use.
  4. Starts the run only after you say yes, follows it, and reports the score, timings, slowest pages and errors.

What the assistant cannot do

These rules are enforced by the UserLimit server, whichever client you use:

  • No run without your confirmation. The server issues a signed summary of the run. A start that doesn't match it, or comes more than 15 minutes later, is refused.
  • Verified domains only. Tests go only to domains verified with DNS or a meta tag, exactly as in the web app.
  • No payments. The assistant can read your balance and estimate a run, but buying hours and plans stays on userlimit.com.
  • Runs stop themselves. An error threshold winds a run down, and you can ask the assistant to stop it any time.

Results are shared as a public report link by default, as with the Share button in the web app. Ask the assistant not to share if you want the report private.

Tools

Clients see 19 tools, each marked read-only or destructive so the client knows when to ask for approval.

Group Tools
Domains get_account_overview, add_domain, verify_domain, remove_domain
Scenarios create_scenario, update_scenario, validate_scenario, get_scenario, delete_scenario
Tests preview_test_settings, create_test, update_test, get_test, get_account_usage, delete_test
Runs prepare_test_run, start_test_run, get_test_run, stop_test_run

If it doesn't connect

  • Sign-in window doesn't open: check that the client supports remote servers with OAuth. Older clients only run local (stdio) servers.
  • "Unauthorized" after a while: tokens last an hour and refresh on their own. If refresh fails, disconnect and connect again.
  • The assistant says a domain isn't verified: verification is the same as in the web app. Check the record with Verify a domain.
  • Out of hours: the assistant explains the limitation and links to the account dashboard. See Usage and billing.
Esc

Documentation

What UserLimit tests Run your first load test Verify a domain Connect an AI assistant (MCP) Sessions, arrivals and concurrency Create and import a scenario Validate and fix a scenario Configure traffic and test settings Start, stop and monitor a run Read a load-test report Test a storefront browsing scenario Test campaign and launch readiness Write a client-facing test summary Playwright compatibility and conversion Metrics and measurement Usage and billing Troubleshoot a failed or misleading test Test production safely Contact support
↑↓ to move to open