oxagen-evals

Benchmarks measuring how the Oxagen ontology layer changes coding-agent performance on real engineering tasks. Compares vanilla Claude Code against Claude Code + Oxagen MCP server across navigation, refactor, and bug-fix scenarios on a typed code graph.

{
  "mcpServers": {
    "oxagen-evals": {
      "command": "uvx",
      "args": [
        "oxagen-evals"
      ]
    }
  }
}

claude mcp add oxagen-evals uvx oxagen-evals

{
  "mcpServers": {
    "oxagen-evals": {
      "command": "uvx",
      "args": [
        "oxagen-evals"
      ]
    }
  }
}