mcpbench

elnukk/mcpbench

★ 1 stars Python 🤖 AI/LLM Updated today

A testing library for MCP servers.

View on GitHub → Try with Claude — $10 free →

Quick Install

Copy the config for your editor. Some servers may need additional setup — check the README.

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "mcpbench": {
      "command": "uvx",
      "args": [
        "mcpbench"
      ]
    }
  }
}

Or install with pip: pip install mcpbench

README Excerpt

A testing library for MCP servers. Write a prompt, declare which tool you expect Claude to call, run it N times, and get a hit rate. When something fails, `--diagnose` asks Claude to explain why the tools are being confused and suggest a fix. --- MCP server developers write tool descriptions blind. You ship, and discover in production that Claude keeps calling `list_issues` when the user says "search for issues" — because `search_issues` and `list_issues` have descriptions that are too similar.

Tools (1)

click

mcpbench

Quick Install

README Excerpt

Tools (1)

Related AI/LLM Servers