Loadster

Loadster MCP

Loadster is a cloud load testing and monitoring platform. Its built-in Model Context Protocol (MCP) server lets AI agents like Claude, ChatGPT, Codex, Cursor, and VS Code write and play load test scripts, build load test scenarios, set up monitors, and analyze results in your Loadster account.

This repository is the public home for connecting to that server: setup snippets for each client, the Claude Code plugin, and the metadata behind Loadster's listings in MCP directories. The server itself is hosted by Loadster at https://api.loadster.com/mcp, and its source is not published here. Loadster is also an official plugin in the ChatGPT and Codex plugin directory.

The maintained, full-length guide is the AI Assistants & MCP chapter of the Loadster manual. This README is the short version.

The Loadster MCP endpoint

OAuth is the easiest way to connect. Clients that support MCP OAuth send you to Loadster to approve the connection in your browser, and you can review or revoke connected agents on the API & MCP page in your Loadster settings. The approval page leads with the web address Loadster can verify for the app, so check that it's one you expect before you approve. You also choose an access level there: Read (selected to start with), Read+Write, or Read+Write+Execute. To change it later, revoke the connection and connect again. Changing or resetting your Loadster password disconnects every OAuth-connected agent, so reconnect each one afterward. Personal access tokens keep working.

Personal access tokens are for clients that can't do OAuth, and for CI jobs and scripted agents. The same kind of token also works for calling the Loadster REST API directly. Create one in the Loadster dashboard under Settings → API & MCP → Personal Access Tokens, choose its access level, copy it right away (it's only shown once), and send it as Authorization: Bearer YOUR_TOKEN. A token acts as you within the team where you created it, so treat it like a password.

Connecting Claude Code to Loadster

The quickest way is the plugin from this repository, which adds the Loadster MCP server plus skills for a sensible load testing workflow: setup to connect, setup-script for a first script, setup-scenario for a first load test scenario, setup-monitor for a first synthetic monitor, load-test for a full test, and results for reading what happened.

/plugin marketplace add loadster/loadster-mcp

/plugin install loadster@loadster

Then run /mcp, pick loadster, and approve the OAuth connection in your browser. Ask Claude something like _"List my Loadster projects"_ to confirm it can reach your account.

Without the plugin, add the server directly and Claude Code will walk you through OAuth:

claude mcp add --transport http loadster https://api.loadster.com/mcp

Or with a personal access token, for headless use:

claude mcp add --transport http loadster https://api.loadster.com/mcp --header "Authorization: Bearer YOUR_TOKEN"

The equivalent project-level .mcp.json:

{

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

Connecting Claude Desktop and claude.ai to Loadster

Claude Desktop and claude.ai connect to remote MCP servers as custom connectors, which use OAuth.

1. Open Settings → Connectors and choose Add custom connector.
2. Enter a name and the URL https://api.loadster.com/mcp.
3. Choose Connect and approve the Loadster OAuth connection when your browser opens.
4. In a conversation, open the tools menu and turn on the Loadster connector.

Connecting ChatGPT and Codex to Loadster

Loadster is an official plugin in the ChatGPT and Codex plugin directory, with the setup, load-test, and results skills. The new setup-script, setup-scenario, and setup-monitor skills reach it when the directory listing is updated.

• ChatGPT: open the Loadster plugin, choose Install plugin, then Connect and approve the Loadster OAuth connection. Start a conversation with @Loadster. On a Business or Enterprise workspace, an admin may need to make it available under Workspace settings → Plugins first.
• Codex: install Loadster from the Plugins tab in the Codex desktop app, or run /plugins in the Codex CLI, then start a new session.

The Codex IDE extension doesn't support plugins, so it connects through Codex's MCP configuration, which the desktop app and CLI share:

codex mcp add loadster --url https://api.loadster.com/mcp

codex mcp login loadster

Or with a personal access token in ~/.codex/config.toml:

[mcp_servers.loadster]

url = "https://api.loadster.com/mcp" bearer_token_env_var = "LOADSTER_MCP_TOKEN"

If the plugin isn't available on your ChatGPT account, you can still add Loadster as a custom connection in Developer mode (Settings → Security and login). Open Plugins, add a connection with the URL https://api.loadster.com/mcp, then add it to a new conversation from the tools menu and approve the OAuth connection.

Connecting Cursor to Loadster

Or add it to .cursor/mcp.json in your project (or ~/.cursor/mcp.json for all projects) and approve the OAuth connection when Cursor first connects:

{

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

To use a personal access token instead, add "headers": { "Authorization": "Bearer ${env:LOADSTER_MCP_TOKEN}" } and set LOADSTER_MCP_TOKEN in the environment before starting Cursor.

Connecting VS Code to Loadster

Or run MCP: Add Server from the Command Palette, or add it to .vscode/mcp.json:

{

"servers": { "loadster": { "type": "http", "url": "https://api.loadster.com/mcp" } } }

To use a personal access token, declare a promptString input with "password": true and reference it in an Authorization header. The manual has the complete example.

Connecting other MCP clients to Loadster

Any client that supports remote servers over Streamable HTTP can connect the same way. For clients that only support local (stdio) servers, the mcp-remote bridge usually works. Leave out the --header arguments to have it run the OAuth flow instead of using a token:

{

"mcpServers": { "loadster": { "command": "npx", "args": ["-y", "mcp-remote", "https://api.loadster.com/mcp", "--header", "Authorization: Bearer YOUR_TOKEN"] } } }

What an agent can do with Loadster

The server exposes most of the Loadster dashboard. Clients fetch the live tool list when they connect, so this table is a map of the surface rather than the source of truth. The MCP tool reference in the Loadster manual is generated from the tool definitions themselves, with each tool's parameters and whether it is a Read, Write, Sensitive Write, or Execute tool.

Scripts, scenarios, datasets, and monitors can carry tags. The list tools report each item's tag names, and the create and update tools take a tags list. Tags are shared across your team, so agents reuse the names you already have and create a new tag only when you ask for one.

Updating a script, scenario, monitor, or dataset changes only the fields the agent sends and keeps the rest. A list that's sent, such as a scenario's bot groups or a dataset's rows, replaces the stored list rather than merging into it. The load test tools take loadTestId and scenario bot groups take engineId; the older names testId and loadEngineId still work in 2.0 and will be removed in the next major version.

The tools an agent sees depend on the access level you grant. Read sees the tools that read, Read+Write adds the tools that create, change, delete, and play scripts, and Read+Write+Execute adds launch_load_test, stop_load_test, and enable_monitor. Launching a load test and enabling a monitor send real traffic and burn Fuel unless your plan covers them, so the server tells the agent to get your go-ahead before either. Before a launch the agent gives you your balance and, when the test would burn Fuel, the Fuel estimate range the dashboard shows. A launch with no estimate runs on your plan and burns no Fuel, and one with an estimate runs entirely on Fuel. If your plan can't run a test right now, the agent's launch is refused and you can launch it on Fuel from the dashboard. Monitors are covered by a plan that includes monitoring up to its limits, and use Fuel beyond them. At the other two levels the agent gets everything ready and hands off so you push the launch button in the dashboard.

Some actions are always left to humans. An agent cannot manage notification policies or maintenance windows, or administer your team, billing, or Fuel.

Every tool carries MCP annotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint), so clients that honor them can prompt before writes and deletes.

Safety practices for AI load testing

• The agent acts in your account.