eamonnfaherty/oh-no-mcp-server

Oh No MCP Server

An MCP (Model Context Protocol) server that provides code performance review capabilities. This server exposes tools and prompts to analyze code for performance issues, bottlenecks, and optimization opportunities.

Features

• Highlighted Text Review: Analyze code snippets selected in your IDE
• Single File Review: Analyze an entire file for performance issues
• Directory Review: Recursively scan and analyze all files in a directory
• Flexible Output: Text output for snippets/files, written reports for directories

Installation

Using uvx (Recommended)

For quick installation and usage:

uvx --from git+https://github.com/eamonnfaherty/oh-no-mcp-server oh-no-mcp

From Source

1. Clone this repository:

git clone https://github.com/eamonnfaherty/oh-no-mcp-server.git
cd oh-no-mcp-server

2. Install dependencies using uv:

make install

For development with all extras:

make install-dev

Usage

As an MCP Server

Run the server using the Makefile:

make run

Or directly with uv using the console script:

uv run oh-no-mcp

Or with uvx (no installation needed):

uvx --from git+https://github.com/eamonnfaherty/oh-no-mcp-server oh-no-mcp

Configuration for Claude Desktop

Add to your Claude Desktop configuration (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

Using uvx with git repository (recommended):

{
  "mcpServers": {
    "oh-no": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/eamonnfaherty/oh-no-mcp-server",
        "oh-no-mcp"
      ]
    }
  }
}

Or using uv from a local clone:

{
  "mcpServers": {
    "oh-no": {
      "command": "uv",
      "args": ["--directory", "/path/to/oh-no-mcp-server", "run", "oh-no-mcp"]
    }
  }
}

Available Tools

oh_no

Returns a prompt for reviewing code performance.

Parameters:

• scope (required): Type of review - "text", "file", or "directory"
• content (required): The code text, file path, or directory path
• output_path (optional): For directories, where to write the report

Examples:

# Review highlighted text
{
  "scope": "text",
  "content": "def slow_function():\n    result = []\n    for i in range(1000):\n        result.append(i)"
}

# Review a single file
{
  "scope": "file",
  "content": "/path/to/your/file.py"
}

# Review a directory
{
  "scope": "directory",
  "content": "/path/to/your/project",
  "output_path": "/path/to/output/performance_report.md"
}

Available Prompts

oh_no

Directly invokes a performance review with the same parameters as the tool.

Arguments:

• scope: "text", "file", or "directory"
• content: The code, file path, or directory path
• output_path: (For directories) Where to write the report

What the Review Includes

The performance review analyzes code for:

1. Performance Bottlenecks: Identifies slow operations and inefficiencies
2. Memory Usage: Highlights potential memory leaks or excessive allocations
3. Algorithm Complexity: Analyzes time and space complexity
4. Optimization Suggestions: Provides actionable improvements
5. Best Practices: Recommends industry-standard patterns

Project Structure

oh-no-mcp-server/
├── pyproject.toml
├── README.md
└── src/
    └── oh_no_mcp_server/
        ├── __init__.py
        └── server.py

Development

This project uses uv for dependency management. All common development tasks are available via the Makefile.

Makefile Commands

Installation

• make install - Install project dependencies using uv sync
• make install-dev - Install dependencies with all extras for development

Code Quality

• make format - Auto-format code using black and fix issues with ruff
• make lint - Check code style with ruff and verify formatting with black (no changes made)

Testing

• make test - Run all unit tests using pytest
• make test-coverage - Run tests with coverage reporting (terminal + HTML)
• make test-coverage-report - Open the HTML coverage report in your browser

Evaluation

• make eval - Run evaluation tests with standard output
• make eval-verbose - Run evaluation tests with detailed output (includes print statements)

Utilities

• make clean - Remove all build artifacts, cache files, and coverage reports
  • Deletes __pycache__ directories
  • Removes .pyc files
  • Cleans up .egg-info directories
  • Removes .pytest_cache, htmlcov, .coverage
  • Removes dist and build directories
• make run - Start the MCP server using uv run oh-no-mcp

Testing Framework

This project includes two types of tests:

Unit Tests (tests/)

Traditional unit tests with 100% code coverage testing individual functions and components.

Evaluation Tests (evals/)

Advanced semantic similarity-based tests that compare MCP tool outputs using:

• BERT embeddings - Semantic understanding via sentence transformers
• Self-BLEU - N-gram overlap measurement
• Cosine similarity - Normalized embedding comparison

Run evaluation tests:

make eval              # Run evaluations
make eval-verbose      # Run with detailed output

Each evaluation provides detailed metrics and a pass/fail result. See for more details.

License

MIT