OASiS

What it does

OASiS provides a unified Model Context Protocol (MCP) interface that connects AI coding agents to eight independent finite element method (FEM) solvers. It allows AI agents to operate solvers, couple different codes across physics domains, and verify simulation correctness through a single protocol, effectively acting as an intelligent bridge for computational mechanics.

Tools

• prepare_simulation: Combines knowledge, examples, and templates for setup.
• run_simulation: Executes Python-based solvers (FEniCS, NGSolve, scikit-fem, DUNE).
• run_with_generator: Generates input and runs compiled solvers (4C, deal.II, Kratos).
• knowledge: Accesses physics knowledge, pitfalls, and cross-backend collation.
• discover: Lists solvers and their capabilities matrix.
• examples: Provides real test files and templates.
• coupled_solve: Handles cross-solver domain decomposition.
• transfer_field: Extracts and transfers fields between solver outputs.
• visualize: Automates field statistics and plots.
• generate_mesh: Creates Gmsh meshes for standard domains.
• developer: Enables source architecture browsing and extension.
• session_insights: Reviews patterns to improve agent knowledge.
• reload_catalog: Hot-reloads backend catalogs.

Installation

{
  "mcpServers": {
    "oasis": {
      "command": "python",
      "args": ["-m", "server"],
      "env": {
        "PYTHONPATH": "src",
        "PYVISTA_OFF_SCREEN": "true"
      }
    }
  }
}

Supported hosts

Confirmed support for Claude Code, Cursor, Windsurf, and GitHub Copilot.