MCP Server MCP

retriEVAL

By github 👁 1 views ▲ 0 votes

LLM evals as MCP tools: score outputs for faithfulness, relevancy, and hallucination.

Homepage Verified source GitHub

Install

oci:ghcr.io/hcarrillo001/retrieval-mcp:1.0.0

Configuration Example

{
  "remotes": [],
  "packages": [
    {
      "registryType": "oci",
      "identifier": "ghcr.io/hcarrillo001/retrieval-mcp:1.0.0",
      "transport": {
        "type": "stdio"
      },
      "environmentVariables": [
        {
          "description": "Judge model API key. Not needed if you set RETRIEVAL_JUDGE_BACKEND=ollama to run the judge locally.",
          "isSecret": true,
          "name": "ANTHROPIC_API_KEY"
        },
        {
          "description": "Which judge to score with: anthropic (default), openai, or ollama for a fully local judge.",
          "default": "anthropic",
          "name": "RETRIEVAL_JUDGE_BACKEND"
        },
        {
          "description": "Judge model id, e.g. claude-sonnet-4-6 or qwen2.5:14b for Ollama.",
          "name": "RETRIEVAL_JUDGE_MODEL"
        },
        {
          "description": "Spend cap for judge calls. Unset means no cap.",
          "name": "RETRIEVAL_BUDGET_USD"
        }
      ]
    }
  ]
}

mcp model-context-protocol oci