rewirebio.iobenchmarks

AgentsRelease 2026-10-22-1b7cf027e4f6

Connect your agent

Let a coding agent query the reviewed evidence and run the benchmarks on your machine, for example to check a published result against your own model or configuration.

Install in Claude Code

The bio-model-benchmarks plugin. The server needs uv.

/plugin marketplace add rewire-bio/rewire-benchmarks
/plugin install bio-model-benchmarks@rewire-benchmarks

How it works

  1. 1

    Install the plugin

    Two commands in Claude Code add a benchmark skill and a read-only server for the database's knowledge graph.

  2. 2

    Ask about the evidence

    The agent finds models, benchmarks, published results and their sources, with review status and caveats.

  3. 3

    Reproduce and compare

    It runs a benchmark with the rewirebench tool on your machine and sets your result beside the published one.

Try asking

  • Which models have published results on the MFASS splicing benchmark, and how do they compare with the baseline?
  • Reproduce the published MFASS-v2 k-mer baseline locally and compare it with the published values.
  • Run my model through the ProteinGym protocol next to the registered baselines and report the difference.

Other ways to connect

Adds just the rewire-benchmark-graph MCP server to Claude Code, without the skill.

claude mcp add rewire-benchmark-graph -- uvx --from git+https://github.com/rewire-bio/rdf-kg-mcp@v0.1.0 rdf-kg-mcp serve https://benchmarks.rewirebio.io/agents/kg.tar.gz

What happens where

  • Runs on your machine

    Your model, weights and data are not sent to Rewire.

  • Nothing submitted for you

    Submitting a result to the database is a separate step that needs your explicit consent.

  • Checked monthly

    Each use case's evidence is re-checked against new publications every month.

  • Reproduce, don't assume

    A source-checked result is a reviewed transcription of a paper's number. Running it yourself is how you find out.

For agents reading this site: plain-text versions at /llms.txt and /agents.md.