mcp-occams-razor

Occam's Razor Thinking Tool is an MCP server extension that guides LLMs through a structured, multi-stage cognitive workflow before implementing code. It enforces stages such as Context Analysis, Outcome Definition, Solution Exploration, Simplicity Evaluation, and Implementation, with a focus on minimizing unnecessary complexity while preserving correct functionality. When integrated as an MCP endpoint, it provides a single, stateless interface that coordinates these stages, encouraging transparency and metacognition in the model's reasoning and ensuring that proposed solutions are simple, maintainable, and well-aligned with user intent. The server exposes a tool interface (occams_razor_thinking) that can be invoked by the hosting LLM to perform the thinking process and return a final, implementable code artifact or design rationale, along with structured feedback at each stage.

Usage typically involves sending a request to the MCP endpoint describing the task or problem domain. The tool then prompts the LLM to articulate its current stage, its planned approach, potential trade-offs, and a minimal viable solution before proceeding to implementation. This process emphasizes minimalism, clarity, and alignment with project constraints, aiming to reduce feature creep and unnecessary complexity in generated code.

Prerequisites:

• Node.js and npm installed on your system (for a Node-based MCP server).
• Access to the MCP hosting environment where you will deploy the server (local or cloud).

Installation steps:

1. Clone or download the MCP Occam's Razor Thinking Tool repository to your environment.
2. Navigate to the project directory.
3. Install dependencies:
   • npm install
4. Start the MCP server:
   • npm run start (If a custom start script is provided, use that command instead, e.g., node path/to/server.js or a docker-based deployment.)
5. Verify the server is listening on the expected port and accessible by your MCP client.

Validation:

• Send a simple test prompt to ensure the multi-stage workflow is triggered and that the response includes stage-wise reasoning and a final implementable result.

Notes and tips:

• Environment variables: adjust MCP_DEBUG for verbose logs during development. Secure any sensitive configuration in your deployment environment.
• If you encounter mismatches between the promised stages and the actual outputs, verify the stage prompts within the MCP interface configuration and ensure the orchestrator is routing stage results correctly.
• This tool emphasizes simplicity and maintainability; if the model proposes overly complex alternatives, re-run with a constraint to minimize features and dependency surface.
• If using a containerized deployment, ensure the container exposes the endpoint securely (HTTPS, auth as needed) and that resource limits (CPU/mmemory) are appropriate for the workload.
• Compatibility: while designed as an MCP endpoint, you may adapt the tool to other orchestration environments as needed, ensuring the same stage semantics are preserved.