NameIntel
Score brand names across domain availability, USPTO trademark conflict, social handle availability (12+ platforms), SEO strength, and AI findability (GEO) in a single call. Pay-per-call via the x402 micropayment protocol (USDC on Base + Solana) — no signup, no API keys.
README
NameIntel — MCP Server
Clear to launch. Score brand names across domain, trademark, social, SEO, and AI findability — pay-per-call via x402 micropayments, no signup, no API keys.
NameIntel is a remote Model Context Protocol server that scores brand names across five dimensions and returns a composite 0–100 score with sub-scores and a verdict. The scoring engine is also reachable as a REST API. Every endpoint is priced in USDC and settled over the x402 protocol on Base and Solana — there are no API keys to provision, no accounts to sign up for, and no rate-limit setup.
This repository is the public face of the project: install instructions, registry submissions, docs, and discovery manifests. The hosted service runs at nameintel.io.
Quick install
Claude Code:
claude mcp add --transport http nameintel https://mcp.nameintel.io/mcp
Claude Desktop: Settings → Connectors → Add → paste https://mcp.nameintel.io/mcp.
That's it. There's no API key. The first time you call a tool the request returns HTTP 402 with the price (between $0.01 and $0.05 USDC depending on the tool); sign the payment with a Base or Solana wallet, retry, and it works. The Coinbase x402 facilitator verifies and settles in under a second.
What it scores
| Dimension | What you get |
|---|---|
| Domain | Availability + pricing across .com, .io, .ai, .app, .co, .dev and other TLDs. Premium and auction detection. |
| Trademark | USPTO conflict screening. Exact matches, similar marks, Nice class analysis, risk level. |
| Social | Handle availability across 12+ platforms (X, Instagram, TikTok, LinkedIn, YouTube, GitHub, Threads, more). |
| SEO | Keyword strength, search-friendliness, and competition risk. |
| GEO | Industry-first. AI findability — how likely the name surfaces correctly in answers from ChatGPT, Perplexity, Claude, and Gemini. Entity collision, semantic distinctiveness, corpus saturation, phonetic clarity. |
Tools
| Tool | Endpoint | Price |
|---|---|---|
score_name |
POST /api/v1/score |
$0.05 USDC |
| Basic name score (no GEO) | POST /api/v1/score/basic |
$0.02 USDC |
check_domain |
POST /api/v1/check/domain |
$0.01 USDC |
check_trademark |
POST /api/v1/check/trademark |
$0.01 USDC |
check_social |
POST /api/v1/check/social |
$0.01 USDC |
check_geo |
POST /api/v1/check/geo |
$0.03 USDC |
REST API
If MCP isn't your fit, the same scoring engine is exposed as a REST API at https://api.nameintel.io. The OpenAPI 3.1 spec is served live at /openapi.json and any framework that auto-imports OpenAPI (LangChain, Mastra, Vercel AI SDK) will wire up tool wrappers from it.
# Calling /api/v1/score without payment returns a 402 with the full menu:
curl -X POST https://api.nameintel.io/api/v1/score \
-H "content-type: application/json" \
-d '{"name": "Lumina"}'
# → HTTP 402, body is a JSON price menu + payment manifest link
Machine-readable manifests
Every important surface is self-describing. Agents and discovery tools should prefer these over scraping prose.
- /.well-known/mcp.json — MCP server URL, tool inventory, prices
- /.well-known/x402 — payment manifest (chains, receivers, facilitators)
- /openapi.json — OpenAPI 3.1 spec for the REST API
- /changelog.json — version + breaking-change tracking
- /health — liveness plus full capability inventory
- llms.txt — LLM-readable summary with embedded YAML manifest
Why this exists
When an AI agent helps a founder start a company, the first thing it needs is a clear-to-launch brand name. Today there is no agent-native tool for this — every existing naming tool requires a human in the loop, an API key, or a subscription. NameIntel is the tool that agent calls: one endpoint, USDC settlement, no friction.
The pricing also makes new use cases viable. A $0.01 trademark check or $0.05 brand-name score can't profitably exist on card rails — Stripe's smallest economically-viable transaction is around $0.50 net of fees. USDC on Base settles those amounts with sub-cent fees and sub-second finality, which is the difference between a market that exists and one that does not.
Repository contents
| Path | What it is |
|---|---|
registry-submissions/ |
Drafted submission packs for MCP / x402 / agent-discovery registries (MCP.so, MACH Alliance, Azure API Center, x402.org, awesome-mcp-servers, Kong, MCP Compass, AAIF). Exact text + submission paths. |
docs/ |
Public-facing strategy and design docs — including agent-discovery-strategy.md, the rationale for the discovery surface this service ships. |
LICENSE |
MIT. |
Status
| Status | |
|---|---|
| Service | ✅ Live at api.nameintel.io and mcp.nameintel.io |
| x402 mode | bypass (free during ecosystem launch). Will flip to enforce after live-payload verification. |
| Phase | 1 — MCP + API. Phase 2 (Stripe web app for human users) lands later. |
Built with
Built by Silverback CTO and CogleGroup using the Velocity Process methodology.
- AWS App Runner + CloudFront + Route 53
- TypeScript + Express + the MCP TypeScript SDK
- Claude (via AWS Bedrock) for the GEO score
- USPTO TSDR API for trademark screening
- Coinbase x402 facilitator for payment verification
Contact
- Issues / suggestions: open an issue on this repo
- Commercial / agency: todd@silverbackcto.com
- Twitter/X: @toddamerrill
License
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。
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。
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。
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。
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。
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
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