mcp-workbench-mcp-server

Expose MCP Workbench capabilities as an MCP server, so AI agents (Claude, Cursor, etc.) can inspect, test, and validate MCP servers programmatically.

Claude Code demo

How it differs from mcp-workbench CLI

The mcp-workbench CLI is designed for human developers running commands in a terminal. This project wraps that CLI as an MCP server, so AI agents can call the same capabilities as structured tools with typed inputs and outputs.

Prerequisites

The mcp-workbench CLI must be installed and available on your PATH:

npm install -g mcp-workbench

Or set the MCP_WORKBENCH_CLI environment variable to the path of the binary.

Installation

npm install -g mcp-workbench-mcp-server

Or clone and build from source:

git clone https://github.com/raeseoklee/mcp-workbench-mcp-server.git
cd mcp-workbench-mcp-server
npm install
npm run build

Usage

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "mcp-workbench": {
      "command": "mcp-workbench-server"
    }
  }
}

Cursor

Add to your MCP server configuration:

{
  "mcpServers": {
    "mcp-workbench": {
      "command": "mcp-workbench-server"
    }
  }
}

Manual

# Run directly (stdio transport)
mcp-workbench-server

# Or via node
node dist/index.js

Available Tools

inspect_server

Connect to an MCP server and inspect its capabilities.

Input:

Output: Server name, version, protocol version, and a capability matrix (tools, resources, prompts, completions, logging).

generate_spec

Auto-generate a YAML test spec by discovering server capabilities.

Input:

Output: The generated YAML spec, test count, and any warnings (e.g. TODO placeholders).

run_spec

Run a test spec against an MCP server.

Input:

*At least one of specText or specPath must be provided.

Output: Full test report with pass/fail counts, duration, and per-test results with assertion details.

explain_failure

Analyze test results and explain failures with actionable recommendations.

Input:

Output: Classified failure causes (auth, placeholder, discovery, protocol, assertion) with counts and recommendations.

Typical Workflow

1. Inspect the server to see what it supports

2. Generate a test spec based on discovered capabilities

3. Run the spec to validate the server

4. Explain any failures to get actionable next steps

Security Considerations

• No tokens or credentials are stored by this server

• Authentication headers are passed per-call and not persisted

• The server spawns mcp-workbench CLI as a subprocess with the current environment

• Spec files written to temp directories are cleaned up after use

MVP Limitations

• Only stdio transport is supported for connecting to this MCP server itself

• No streaming of test results (waits for full completion)

• The headers field on run_spec is reserved for future use; headers are defined in the spec YAML

• No caching of inspection or generation results between calls

License

Apache-2.0