recursive-companion-mcp

democratize-technology/recursive-companion-mcp

3.4

If you are the rightful owner of recursive-companion-mcp and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

The Recursive Companion MCP server implements iterative refinement through self-critique cycles, enhancing response quality by leveraging incremental processing and domain-specific optimizations.

The Recursive Companion MCP server is designed to enhance the quality of model-generated responses through a structured process of iterative refinement. This process involves generating an initial draft, critiquing it using multiple parallel critiques, revising the response based on these critiques, and measuring convergence using cosine similarity. The server is optimized for various domains, including technical, marketing, strategy, legal, and financial, ensuring that the responses are tailored to specific contexts. By breaking down the refinement process into discrete steps, the server avoids timeouts and provides progress visibility, allowing for real-time updates in user interfaces. Additionally, the server supports multiple concurrent refinement sessions, making it scalable and efficient for diverse applications.

Features

  • Incremental Refinement: Breaks down the refinement process into discrete steps to avoid timeouts.
  • Mathematical Convergence: Uses cosine similarity to determine when the refinement process is complete.
  • Domain-Specific Optimization: Automatically detects and optimizes for various domains such as technical, marketing, and legal.
  • Progress Visibility: Provides immediate feedback at each step for real-time UI updates.
  • Parallel Sessions: Supports multiple concurrent refinement sessions for scalability.

Tools

  1. start_refinement

    Initiates a refinement session for a given input.

  2. continue_refinement

    Continues the refinement process for an existing session.

  3. get_final_result

    Retrieves the final refined result for a session.

  4. get_refinement_status

    Checks the progress of a refinement session without advancing it.

  5. list_refinement_sessions

    Lists all active refinement sessions.