not-a-feature/evo2-mcp

evo2-mcp

The evo2-mcp server exposes Evo 2 as a Model Context Protocol (MCP) server, providing tools for genomic sequence analysis. Any MCP-compatible client can use these tools to score, embed, and generate DNA sequences.

Features

• Sequence Scoring: Compute log probabilities for DNA sequences
• Sequence Embedding: Extract learned representations from intermediate model layers
• Sequence Generation: Generate novel DNA sequences with controlled sampling
• Variant Effect Prediction: Score SNP mutations for variant prioritization
• Multiple Model Checkpoints: Support for 7B, 40B, and 1B parameter models

Getting Started

Prerequisites: Python 3.12

1. Install Evo2 dependencies: See Installation Guide for details.

   conda install -c nvidia cuda-nvcc cuda-cudart-dev
   conda install -c conda-forge transformer-engine-torch=2.3.0
   pip install flash-attn==2.8.0.post2 --no-build-isolation
   pip install evo2
   

2. Install evo2-mcp:

   pip install evo2-mcp
   

3. Activate MCP Server: Add the following to your mcp.json configuration:

   {
   "mcpServers": {
      "evo2-mcp": {
         "command": "python",
         "args": ["-m", "evo2_mcp.main"]
      }
   }
   }
   

For detailed installation instructions, see the Installation Guide.

Usage

Once installed, the server can be accessed by any MCP-compatible client. For available tools and usage examples, see the Tools Documentation.

Available Tools

• score_sequence - Evaluate DNA sequence likelihood
• embed_sequence - Extract feature representations
• generate_sequence - Generate novel DNA sequences
• score_snp - Predict variant effects
• get_embedding_layers - List available embedding layers
• list_available_checkpoints - Show supported model checkpoints

See the Tools Documentation for detailed API reference and examples.

Documentation

• Installation Guide - Detailed installation instructions
• Tools Reference - Complete API documentation and usage examples
• Development Guide - Contributing and testing information
• Changelog - Version history and updates

You can also find this project on BioContextAI, the community hub for biomedical MCP servers.

Citation

If you use evo2-mcp in your research, please cite:

@software{evo2_mcp,
  author = {Kreuer, Jules},
  title = {evo2-mcp: MCP server for Evo 2 genomic sequence operations},
  year = {2025},
  url = {https://github.com/not-a-feature/evo2-mcp},
  version = {0.2.2}
}

For the underlying Evo 2 model, please also cite the original Evo 2 publication.

License and Attribution

The banner image in this repository is a modified version of the original Evo 2 banner from the Evo 2 project, which is released under the Apache 2.0 License. It was modified using Google Gemini "Nanobana" and GIMP.