evo

The Evo MCP server exposes Evo SDK functionality to MCP clients via a standard, self-hosted interface. It acts as a bridge between your AI models or agents and Seequent Evo capabilities, such as workspaces, geoscience objects, and block models. Clients like VS Code with MCP support, Cursor, Claude Desktop, or other MCP-enabled tools can connect through STDIO or a streamable HTTP transport to access Evo APIs through the server. Tools are modular and registered based on the MCP_TOOL_FILTER setting, and the EvoContext handles OAuth tokens and lazy initialization of the Evo SDK client when a tool requires access.

Once running, you can invoke registered tools to create, list, or manipulate Evo resources (workspaces, objects, and block models) from within your AI workflow. The server is designed to be extensible, allowing you to add your own tools that leverage Evo SDK capabilities. For typical usage, connect your MCP client, authenticate if prompted, and start issuing tool calls to manage workspaces, generate structured geoscience objects from CSV data, and perform data workflows inside Evo from your AI agents.

Prerequisites

• Python 3.10 or newer
• Access to Seequent Evo (https://evo.seequent.com)
• Git installed on your machine

Installation steps

1. Clone the Evo MCP repository

git clone https://github.com/SeequentEvo/evo-mcp.git

2. Navigate to the repository root

cd evo-mcp

3. Set up a Python environment (recommended: uv). If you don’t have uv installed yet, install it from https://docs.astral.sh/uv/

# Example using uv (recommended)
uv sync

4. Install the package in editable mode (this assumes a proper setup.py / pyproject.toml in the repo)

pip install -e .

5. Configure environment variables for Evo access (example placeholders shown; replace with your actual values):

export EVOSDK_CLIENT_ID="your-client-id"
export EVOSDK_CLIENT_SECRET="your-client-secret"
export EVO_ORG="your-organization-id"

6. Run the MCP server via uv (assuming the package name is evo-mcp and it exposes a CLI entry point). If your setup requires a specific module, adjust accordingly:

uvx evo-mcp

7. If you prefer STDIO transport (default) ensure your MCP client is configured to spawn the server process and communicate via stdin/stdout, or switch to a streamable HTTP transport if remote access is required.

Prerequisites recap

• Python 3.10+
• Access to Evo (credentials and org access)
• A compatible MCP client (VS Code, Cursor, Claude Desktop, etc.)

Note on transport: Evo MCP supports two transports – STDIO and streamable HTTP. STDIO is typically recommended for local integrations with-dev tooling like VS Code and Cursor, as it provides lower latency and simpler client configuration. If you need remote access or testing, consider enabling streamable HTTP transport. Tools are loaded based on MCP_TOOL_FILTER; you can customize which Evo capabilities are exposed by setting this filter. Ensure OAuth tokens are cached and refreshed as needed; EvoContext handles this lazily on first tool invocation. If you encounter connection issues, verify that your environment variables for Evo (client ID/secret and org) are correct and that network access to Evo services is allowed from your host.