hernaninverso/eleata-verify-mcp

Overview

An MCP server that checks a claim against provided evidence and returns a verdict of Supported, Refuted, or NEE (Not Enough Evidence). It acts as a grounding or hallucination guard for AI agents by validating factual assertions.

Best for

Best for
Developers building AI agents that need a lightweight, rule-based fact-checking step against known sources.

Use cases

• Verify an AI agent's output against source documents before presenting to users
• Check a user-submitted claim against a knowledge base for fact-checking
• Add a validation step in a retrieval-augmented generation pipeline

How to use

Tools exposed

• verify_claim
• check_groundedness
• verify_strict

Example client config

{\n  "mcpServers": {\n    "eleata-verify": {\n      "command": "npx",\n      "args": ["-y", "eleata-verify-mcp"],\n      "env": { "EVERIFY_API_KEY": "your_key" }\n    }\n  }\n}

Notes

An MCP server that checks a claim against provided evidence and returns a verdict of Supported, Refuted, or NEE (Not Enough Evidence). It acts as a grounding or hallucination guard for AI agents by validating factual assertions.

0 stars on GitHub. Last updated 2026-06-30.

Use cases

• Verify an AI agent’s output against source documents before presenting to users
• Check a user-submitted claim against a knowledge base for fact-checking
• Add a validation step in a retrieval-augmented generation pipeline

Pros

• Simple, focused API with clear three-state verdicts
• Lightweight JavaScript implementation easy to integrate into MCP-based workflows
• Directly addresses hallucination risk in agentic systems

Cons

• No stars or community adoption yet, so reliability is unproven
• Requires users to supply evidence, limiting use to scenarios with available source material
• No built-in evidence retrieval or search, only verification against provided text

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.