hallucination_confidence_meter

Active

declared in 0.2.0

Evaluates the likelihood of hallucination in LLM responses by comparing against HuggingFace model confidence scores. Designed for risk assessment personas to quantify response reliability. Accepts text snippets or model outputs, returns confidence metrics and potential hallucination warnings. Cross-references with top-performing models from the HuggingFace leaderboard.

Parameters schema

{
  "type": "object",
  "required": [
    "text"
  ],
  "properties": {
    "text": {
      "type": "string",
      "description": "The LLM-generated text to evaluate for hallucination risk"
    },
    "async": {
      "type": "boolean",
      "description": "If true, returns a job_id immediately (<200ms) instead of waiting for the result. Poll the result with job_result(job_id). Use for slow tools to avoid client timeouts."
    },
    "model_id": {
      "type": "string",
      "description": "Optional specific HuggingFace model ID to use for evaluation"
    },
    "threshold": {
      "type": "number",
      "default": 0.7,
      "maximum": 1,
      "minimum": 0,
      "description": "Confidence threshold below which hallucination warnings are triggered"
    }
  }
}

What this tool wraps· 0 endpoints

min confidence0.70 0.50

No endpoints wrapped at confidence ≥ 0.70.

Parent server

@gapup/mcp-knowledge

https://github.com/getgapup/gapup-mcp

2/7 registries

View full server →