evo2-mcp

MCP server for generating, scoreing and embedding genomic sequences using Evo 2

Getting started

Please refer to the documentation, in particular, the API documentation.

You can also find the project on BioContextAI, the community-hub for biomedical MCP servers: evo2-mcp on BioContextAI.

Installation

Prerequisites

You need to have Python 3.12 or newer installed on your system. If you don't have Python installed, we recommend installing uv.

Installing Evo2 Dependencies (Required)

Important: Evo2 has specific installation requirements that must be completed before installing this MCP server. Follow these steps in order:

1. Install CUDA dependencies (using conda):

   conda install -c nvidia cuda-nvcc cuda-cudart-dev
   conda install -c conda-forge transformer-engine-torch=2.3.0

2. Install flash-attn:

   pip install flash-attn==2.8.0.post2 --no-build-isolation

3. Install evo2:

   pip install evo2

This installation order is strongly recommended to ensure all dependencies are properly configured.

Installing evo2-mcp

Once Evo2 is installed, there are several options to install evo2-mcp:

1. Use uvx to run it immediately:

uvx evo2_mcp

2. Include it in one of various clients that supports the mcp.json standard, please use:

{
  "mcpServers": {
    "server-name": {
      "command": "uvx",
      "args": ["evo2_mcp"],
      "env": {
        "UV_PYTHON": "3.12"
      }
    }
  }
}

3. Install it through pip:

pip install --user evo2_mcp

4. Install the latest development version:

pip install git+https://github.com/not-a-feature/evo2-mcp.git@main

Development and Testing

Using the Dummy Implementation

For testing and development without requiring the full Evo2 model dependencies, you can use a dummy implementation that mimics the Evo2 interface:

export EVO2_MCP_USE_DUMMY=true  # On Linux/macOS
# or
set EVO2_MCP_USE_DUMMY=true     # On Windows (cmd)
# or
$env:EVO2_MCP_USE_DUMMY="true"  # On Windows (PowerShell)

This is automatically enabled in GitHub Actions CI/CD pipelines to speed up testing without requiring access to actual model weights.

The dummy implementation:

• Returns realistic output shapes and types
• Generates plausible random values for scores and embeddings
• Does not require downloading or loading any model weights
• Is deterministic (uses fixed random seed) for reproducible tests

Running Tests

To run tests with the dummy implementation:

EVO2_MCP_USE_DUMMY=true pytest

To run tests with the real Evo2 model (requires model installation):

pytest -m real_evo2

Contact

If you found a bug, please use the issue tracker.

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 = {0.1.0}
}

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.