我为人工智能代理构建了一个身份图谱——超过3.3亿条经过验证的记录。打破API的限制。
AI代理会对B2B数据产生幻觉，因为它们没有真实的基础数据。我构建了一个。
Nopp的实体图从NPPES从业者数据、40多个州的许可委员会、ODO公司注册和监管文件中提取数据。没有抓取，没有推断。每个实体都经过确定性验证。
我希望你进行压力测试的API接口：
实体图
- POST /api/v1/identify — 名称、NPI或LinkedIn网址 → 验证的标准实体（10 IC）
- GET /api/v1/people/:id — 获取已解析的个人记录（3 IC）
- POST /api/v1/people/search — 按州、城市、资质筛选120万+持证专业人士（10 IC）
- GET /api/v1/organizations/:id — 获取组织记录（3 IC）
智能
- POST /api/v1/company/enrich — 实时增强，优雅回退，绝不返回空数据（10 IC）
- POST /api/v1/contact/enrich — 联系人级别的深度增强（3 IC）
信号
- POST /api/v1/signals/hiring — 检测招聘激增（15 IC）
- POST /api/v1/signals/funding — 融资轮次检测（15 IC）
- POST /api/v1/signals/intent — 竞争对手挫折 + 意图信号（20 IC）
代理
- POST /api/v1/agent/launch/companies — 一次调用实现自主研究 + 外展（50 IC）
此外，还作为MCP服务器发布 — 直接连接Claude、ChatGPT或任何MCP兼容的代理。无需API调用，无需繁琐的设置。
我希望你做的：
- 输入模糊或冲突的数据进行测试 — 同一个人，两个LinkedIn网址，不同的姓名拼写
- 尝试让/company/enrich返回空或错误的数据
- 针对你非常了解的公司测试/signals/intent — 它能捕捉到真实信号吗？
- 告诉我哪里有架构错误、文档不足或使用起来让人烦恼的地方
找到可重复的bug或真实的架构问题 → 我会给你免费的智能积分。没有上限。
nopp.us/api-docs | nopp.us/developers
我是创始人。我会亲自回复每一个技术评论。
查看原文
AI agents hallucinate B2B data because they have no ground truth. I built one.
Nopp's Entity Graph pulls from NPPES practitioner data, 40+ state licensing boards, ODO corporate registries, and regulatory filings. Nothing scraped. Nothing inferred. Every entity is deterministically verified.
The API surface I want you to stress-test:
Entity Graph<p>POST /api/v1/identify — name, NPI, or LinkedIn URL → verified canonical entity (10 IC)
GET /api/v1/people/:id — fetch a resolved person record (3 IC)
POST /api/v1/people/search — filter 12M+ licensed professionals by state, city, credential (10 IC)
GET /api/v1/organizations/:id — fetch an org record (3 IC)<p>Intelligence<p>POST /api/v1/company/enrich — live enrichment, falls back gracefully, never returns empty (10 IC)
POST /api/v1/contact/enrich — contact-level deep enrichment (3 IC)<p>Signals<p>POST /api/v1/signals/hiring — detect hiring surges (15 IC)
POST /api/v1/signals/funding — funding round detection (15 IC)
POST /api/v1/signals/intent — competitor frustration + intent signals (20 IC)<p>Agents<p>POST /api/v1/agent/launch/companies — autonomous research + outreach in one call (50 IC)<p>Also ships as an MCP Server — connect Claude, ChatGPT, or any MCP-compatible agent directly. No API calls, no plumbing.
What I want from you:<p>Feed /identify ambiguous or conflicting inputs — same person, two LinkedIn URLs, different name spellings
Try to get /company/enrich to return empty or wrong data
Test /signals/intent against a company you know well — does it catch real signals?
Tell me where the schema is wrong, underdocumented, or just annoying to work with<p>Find a reproducible bug or a real schema issue → I'll give you free Intelligence Credits. No cap.
nopp.us/api-docs | nopp.us/developers
I'm the founder. I'll reply to every technical comment personally.