Qiniso
The deterministic fact-verification layer for AI agents. Validates the structured facts an agent emits — IBANs, payment cards, VAT and national tax IDs, crypto and bank addresses, domains, emails, phone numbers, securities and academic identifiers, plus dates, currencies and holidays — against checksums and curated authoritative data, not guesses.
README
<div align="center">
<img src="docs/logo.svg" width="96" height="96" alt="Qiniso" />
Qiniso
The deterministic fact-verification layer for AI agents.
Verified, trustworthy data tools for AI agents. "Qiniso" means "truth" in Zulu.
Website · MCP endpoint · MCP Registry
</div>
Agents confidently emit IBANs, phone numbers, domains, VAT numbers and crypto addresses that are subtly — and silently — wrong. Qiniso checks the structured facts an agent produces against checksums and curated authoritative data, so a bad value is caught instead of trusted.
It's the deterministic complement to the guardrail stack: security guardrails check whether output is safe, structure guardrails check it's well-formed, and hallucination guardrails ask another LLM if it's faithful to the prompt. None of them check whether a structured fact is actually correct against the real world — because that needs computation or curated data, not another model's opinion. That's Qiniso.
On arbitrary identifiers, a frontier LLM validates them wrong ~91% of the time, cold and silently. Qiniso: 0%.
Add it to Claude
Settings → Connectors → Add custom connector, and paste — no login, no key:
https://qiniso.qinisolabs.workers.dev/mcp
Stateless, reads no user data, requires no secrets.
Use it as a library
Every check is also a typed function — no MCP required:
npm i @qinisolabs/qiniso
import { validateIban, validateVat } from "@qinisolabs/qiniso";
validateIban("GB82 WEST 1234 5698 7654 32");
// { valid: true, country: "United Kingdom", ... }
What it verifies — 34 tools across 8 domains
| Domain | Tools |
|---|---|
| Identifiers | IBAN, payment card (Luhn + brand), ISBN-13, VIN, GTIN/UPC/EAN barcodes (+ GS1 country) |
| Web / network | TLD & domain (IANA root zone), IP, UUID, URL, email |
| Finance | ISIN, CUSIP, SEDOL, LEI, US ABA routing |
| Crypto | Ethereum (EIP-55), Bitcoin (Base58Check / Bech32) addresses |
| National & tax IDs | Brazil CPF/CNPJ, South Africa ID, Spain DNI/NIE, India Aadhaar, EU/UK VAT |
| Academic | ISBN-10, ISSN, ORCID |
| Locale | Phone (global), date parsing, currency, holidays (~200 countries), UK VAT-by-date |
| Addresses | UK/US address parsing |
Each tool wraps an authoritative method — a published checksum standard, an audited library (libphonenumber-js, jsvat, date-holidays, @noble/hashes), or curated reference data (the IANA root zone, UK VAT history).
What it is not
- Not a live-data provider. It verifies facts you give it; it does not return the current time, weather, or live exchange rates.
- Not a credential sink. It never asks for secrets or API keys. (For JWT signature verification, use a library in your own runtime so the secret never leaves your machine.)
- Not a registration check. It validates a VAT number's checksum, not whether it is live-registered (VIES); it confirms a domain's TLD is real, not that the domain is registered.
Architecture
A TypeScript monorepo. Each domain is a typed library in packages/*; the qiniso umbrella aggregates them and exposes one MCP server over three transports — stdio (local / npx), Streamable HTTP (self-host), and a Cloudflare Worker (the hosted edge endpoint). The same core powers the importable library.
npm install
npm run build
npm test
License
Apache-2.0
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。