Skip to content

New benchmarking engine #250

Description

@voorhs

Hi! Over the past couple of months, I’ve been working on a reimplementation inspired by your benchmark:

https://github.com/deeppavlov/mcp-evals

My goals were:

  • to make it easier to customize and use by building it around the Pydantic-AI framework for LLM agents
  • to apply modern development practices, including enabling safer workflows where LLM agents can write and review benchmark tasks
  • to preserve and build upon the original benchmark content

So far, I’ve recreated two domains from MCPMark: file system and PostgreSQL.

If this sounds interesting, I’d really appreciate your feedback. I’d also be glad to explore possible collaboration or alignment with your work.

Thanks for creating and sharing this benchmark—it’s been very inspiring!

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions