Loadster MCP Server已验证已精选

What is Loadster?

Loadster is a cloud load testing and monitoring platform. You write scripts as HTTP protocol scripts, Browser Bot scripts, or Playwright scripts, run them at scale from cloud engines around the world, and use the same scripts for uptime and performance monitoring.

The Loadster MCP server is built into the platform and hosted at https://api.loadster.com/mcp. It lets AI agents such as Claude, ChatGPT, Codex, Cursor, and VS Code work inside your Loadster account: writing and validating scripts, playing them to check they work, building load test scenarios, setting up monitors, and analyzing load test reports and monitoring results. The agent prepares everything and hands off to you for the parts that should stay human, like launching a full load test.

How do I use Loadster?

The server is remote (Streamable HTTP), so there is nothing to install or run locally. Connect any MCP client to https://api.loadster.com/mcp and authenticate with OAuth 2.1, or with an MCP token from Settings → AI Agents → MCP Tokens in your Loadster dashboard for CI jobs and clients that can't do OAuth.

Claude Code (plugin, includes skills for a sensible load testing workflow): /plugin marketplace add loadster/loadster-mcp /plugin install loadster@loadster

Or add the server directly: claude mcp add --transport http loadster https://api.loadster.com/mcp

Cursor, VS Code, Codex, and other clients use the same URL. A generic config looks like:

{
  "mcpServers": {
    "loadster": {
      "url": "https://api.loadster.com/mcp"
    }
  }
}

Clients that only support local stdio servers can bridge with mcp-remote. Full per-client instructions, including token-based setups, are in the AI Agents chapter of the Loadster manual and the GitHub repository.

## Key features of Loadster

* Script authoring and validation. Create, update, import, and validate HTTP, Browser Bot, and Playwright scripts, with built-in documentation, examples, and command schemas so the agent writes correct scripts the first time.
* Play before you scale. Play a script as a single bot, watch its status, inspect step results and screenshots, and fix problems before anyone launches a load test.
* Scenario building. Compose load test scenarios from scripts, datasets, and cloud engines in your chosen regions.
* Load test analysis. List past load tests, pull full reports, and annotate them with notes.
* Monitoring. Create and update monitors, review cycle history and summaries, and investigate incidents.
* Script revisions and datasets. Every script change creates a revision you can restore; datasets can be created and appended to for data-driven tests.
* Human in the loop by design. Agents cannot launch or stop a full load test, enable a monitor, or touch billing, team administration, or notification policies.
* MCP annotations on every tool. readOnlyHint, destructiveHint, idempotentHint, and openWorldHint let clients prompt before writes and deletes.
* OAuth 2.1 or scoped tokens. Review and revoke connected agents from the AI Agents page in your Loadster settings.

## Use cases of Loadster
* Ask your coding agent to write a load test script for the API or web app you're building, play it to confirm it works, and set up a scenario for you to launch.
* Turn an existing Playwright test or HAR capture into a Loadster script without leaving your editor.
* Have an agent read a finished load test report and explain where response times degraded and which requests failed.
* Set up uptime and performance monitors for a new service as part of shipping it.
* Investigate a monitoring incident by pulling cycle details and comparing them against recent history.
* Generate or extend datasets for data-driven testing.