eleata-verify-mcp is an MCP (Model Context Protocol) server positioned as a grounding / hallucination guardrail for AI agents. Its core purpose is fact-checking: before an agent is about to trust or repeat a statement, it first compares that statement against its source evidence.
[Main capabilities] The server receives a claim and supporting evidence, and returns a judgment of that evidence regarding the claim: Supported, Refuted, or NEE (Not Enough Evidence). This is a judgment based on natural-language inference, and it also returns a confidence along with an abstained flag for low-confidence cases. The recommended usage is: treat abstained=true, or any conclusion other than Supported, as "do not rely." It can be used as a fact-check / RAG hallucination guard component.
[How to use] This tool wraps the hosted eleata Claim Verifier service and requires obtaining a key first. The officially provided address for obtaining one is https://eleata.io/checkout?p=verifypro（付费渠道），密钥以 sent via Authorization: Bearer. If using the RapidAPI marketplace channel, set EVERIFY_RAPIDAPI=1 and set EVERIFY_API_BASE to the RapidAPI host; in this case the key is sent via X-RapidAPI-Key. The verify operation sends your claim and evidence to the hosted API; see the privacy policy at https://eleata.io/privacy/。
[Technical notes and limitations] The underlying engine is the multilingual NLI model mDeBERTa-v3-mnli-xnli. The official documentation explicitly notes: on public channels, confidence is uncalibrated and does not represent a probability, so you should rely on the verdict and the abstained flag rather than the raw numeric value. This tool's judgment depends on the evidence provided, and when there is insufficient evidence it returns NEE. The materials do not provide specific installation commands, code examples, free quotas, pricing, or platform support scope; these are not yet provided and are subject to the official documentation.
[Applicable scenarios] Suitable for hallucination protection for AI agents, fact verification in RAG pipelines, and any automated scenario that needs to verify statements against sources before accepting or repeating them.
