mcp-prompt-optimizer

The MCP Prompt Optimizer server provides research-backed prompt enhancement capabilities designed to squeeze higher quality responses from language models. It exposes a suite of core tools for analyzing, optimizing, and templating prompts, along with advanced strategies like Tree of Thoughts, Constitutional AI, and Medprompt to improve accuracy, safety, and usefulness across a variety of domains. Users interact with the server by issuing structured tool commands or natural-language prompts that specify the desired optimization strategy, the target prompt, and any domain templates to apply. Typical workflows include analyzing a prompt for clarity and bias, applying a chosen optimization technique, and retrieving domain-specific templates to guide output in areas such as business analysis, product management, or software development.

Core tools include: analyze_prompt (to assess prompt quality and identify issues), optimize_prompt (to apply specific strategies like clarity, specificity, or chain-of-thought), auto_optimize (to automatically select among strategies for best overall improvement), and get_prompt_template (to fetch basic domain-agnostic templates). Advanced tools extend functionality with advanced_optimize (research-backed strategies), get_domain_template (domain-specific templates), and list_domain_templates (to enumerate available templates by domain). These tools enable structured, repeatable prompt improvement workflows suitable for research, development, or production use cases.

Prerequisites:

• Python 3.8 or newer installed on your system
• Access to the repository containing mcp-prompt-optimizer
• (Optional) Claude Desktop setup if you plan to integrate with Claude-based workflows

Installation steps:

1. Clone the repository:

git clone <repository-url>
cd mcp-prompt-optimizer

2. Create and activate a virtual environment (recommended):

python3 -m venv venv
source venv/bin/activate  # On Windows: venv\\Scripts\\activate

3. Install dependencies via the provided script or manually:

./install.sh

Or, install manually:

pip install -r requirements.txt

4. If you plan to run in interactive mode with Claude Desktop, configure interactive setup:

python3 setup_interactive.py

5. Start the MCP server using the standard Python entry point (example path shown in config):

python3 /path/to/mcp-prompt-optimizer/prompt_optimizer.py

6. Optional: add a Claude Desktop configuration snippet to point to the server (see configuration example in README):

{
  "mcpServers": {
    "prompt-optimizer": {
      "command": "python3",
      "args": ["/path/to/mcp-prompt-optimizer/prompt_optimizer.py"],
      "env": {}
    }
  }
}

Tips and common issues:

• Ensure Python 3.8+ is installed and your environment is using the correct python executable when starting the server.
• If commands aren’t recognized, verify the configuration file location and the paths to the server script.
• Use the provided setup_interactive.py for an easier integration with Claude Desktop; this configures prompts and templates automatically.
• When testing, you can run the server directly with: python3 prompt_optimizer.py and enable verbose logging by setting MCP_LOG_LEVEL=debug.
• If you update dependencies, re-run the installation script or pip install -r requirements.txt to ensure compatibility with the latest changes.
• Review available tools and templates via the in-app or CLI commands to tailor prompts to specific domains (Business Analysis, Product Management, Development, etc.).