MCP Setup

Connect LLMTest to your IDE as an MCP server. Works with any tool that supports the Model Context Protocol.

Claude Code

1. Open Claude Code in your terminal
2. Type /mcp and press Enter
3. Choose Add MCP Server
4. Name: llmtest
5. Command:

LLMTEST_API_KEY=your-llmt_-key-here npx -y llmtest-mcp

Cursor, Windsurf, Cline, Roo Code

Add this to your MCP config file:

{
  "mcpServers": {
    "llmtest": {
      "command": "npx",
      "args": ["-y", "llmtest-mcp"],
      "env": {
        "LLMTEST_API_KEY": "your-llmt_-key-here"
      }
    }
  }
}

Config file locations:

• Cursor: ~/.cursor/mcp.json
• Windsurf: ~/.windsurf/mcp.json
• Cline/Roo: Check your extension settings for MCP config

Any MCP-compatible tool

Use the same config above. The MCP server is published as llmtest-mcp on npm.

Verify it works

Once connected, ask your AI assistant:

You should see your flows, call count, and credit balance.

What you can do

• "Find cheaper models for my AI calls" — runs benchmarks and suggests alternatives
• "What AI models am I using and how much do they cost?" — shows usage per flow
• "Route my AI calls through LLMTest" — your assistant updates your code
• "Benchmark my support-bot flow, optimize for cost" — goal-driven benchmark

See MCP Tools Reference for the full list of available tools.