namescope
MCP server for evaluating software project name availability across package registries, GitHub, domains, and the command line, with scoring and evidence.
README
NameScope
Picking a project name is easy. Finding out whether that name is usable across package registries, GitHub, domains, and the command line is the tedious part.
NameScope does that research from your terminal. Give it a project description or a shortlist of names and it returns a scored report with the evidence behind every result. Generation and scoring run locally; network checks use public provider endpoints.
NameScope is built for software projects. It cannot promise that a business name, domain, social handle, or trademark is legally available.
Quick start
NameScope requires Node.js 24 LTS or newer. You can run it without installing anything globally:
npx namescope find "a fast open-source database migration tool"
npx namescope check taskforge
npx namescope rank taskforge taskmint orbitdesk
No account, subscription, payment, hosted backend, Docker setup, or language model is required. A GITHUB_TOKEN is optional and only increases GitHub API rate limits.
How it works
find starts with a project description and generates deterministic candidates. check starts with one name, while rank compares several. All three eventually use the same checking and scoring engine.
flowchart TD
A["Project description or candidate names"] --> B{"Command"}
B -->|find| C["Generate candidates locally"]
B -->|check or rank| D["Use supplied names"]
C --> E["Normalize query variants"]
D --> E
E --> F["Run local quality and CLI checks"]
E --> G{"Offline mode?"}
G -->|Yes| H["Skip remote requests and record unknowns"]
G -->|No| I["Query public providers in parallel"]
I --> J["RDAP, npm, PyPI, crates.io, and GitHub"]
F --> K["Collect evidence, confidence, and timestamps"]
H --> K
J --> K
K --> L["Add optional trademark links and disclaimer"]
L --> M["Calculate weighted dimension scores"]
M --> N{"Blocking conflict?"}
N -->|Yes| O["Cap or block the recommendation"]
N -->|No| P["Keep the composite score"]
O --> Q["Return human, JSON, Markdown, or MCP output"]
P --> Q
Provider-specific commands such as domains and packages return their evidence directly instead of building a complete composite report.
1. Generate or accept candidates
Name generation is local and reproducible. NameScope combines useful words from the description with related terms, compounds, prefixes, suffixes, and blends. Generated names are only ideas; they do not carry availability claims until checks run.
2. Check public and local namespaces
NameScope checks each candidate against the sources relevant to software projects:
| Area | What is checked | Important limitation |
|---|---|---|
| Packages | npm name validity, then exact and separator variants on npm, PyPI, and crates.io | Registry state can change immediately |
| GitHub | Repository-name search and exact user or organization namespace | Search is not proof of global availability |
| Domains | Selected TLDs through public RDAP | A 404 is a useful signal, not a purchase guarantee |
| CLI | Local PATH, common commands, language tools, and shell-reserved names |
Results describe the current machine |
| Name quality | Length, word count, pronunciation proxy, punctuation, numbers, ambiguity, and relevance | These are heuristics, not user research |
| Trademark | Official manual-search links and a legal disclaimer when requested | This is not trademark clearance or legal advice |
Search distinctiveness stays neutral until a search adapter exists. Social handles are not checked in version 0.1.
3. Keep uncertainty visible
A timeout, rate limit, blocked request, unsupported provider, or malformed response remains unknown, unsupported, or error. NameScope never turns a failed lookup into “available.”
Every completed report includes provider evidence, confidence, timestamps, warnings, unknown checks, dimension scores, weights, and a plain-language explanation.
4. Score and rank
The default score combines package uniqueness, GitHub uniqueness, domain options, deterministic name quality, search distinctiveness, and CLI usability. Weights are configurable and normalized by their actual sum, so they do not need to total 100.
A package collision or another confirmed blocking conflict can cap or block a recommendation. A high score still means “strong candidate based on these checks,” not “safe everywhere.”
Commands
| Command | Use it to |
|---|---|
namescope find <description> |
Generate, check, and rank project-name ideas |
namescope check <name> |
Build a complete report for one name |
namescope rank <name...> |
Compare two or more existing candidates |
namescope domains <name> |
Check selected domain extensions only |
namescope packages <name> |
Check selected package registries only |
namescope mcp |
Start the local MCP stdio server |
Useful examples:
# Check selected domains
npx namescope domains taskforge --tlds com,io,dev,app
# Check selected package registries
npx namescope packages taskforge --registries npm,pypi,crates
# Return structured output
npx namescope check taskforge --json
# Prevent every network request
npx namescope check taskforge --offline
# Include preliminary trademark resources
npx namescope check taskforge --include-trademark --nice-classes 9,42
Output formats are human-readable text, --json, --compact-json, and --markdown.
Privacy and offline use
Generation, normalization, quality analysis, scoring, cache handling, and CLI collision checks run locally. Remote commands send only the requested candidate and query variant to the selected providers.
Use --offline to prevent all network requests. NameScope has no telemetry and does not collect project names or descriptions. Read the privacy documentation for provider-by-provider details.
Configuration
Copy namecheck.config.example.json to namecheck.config.json. Use it to change TLDs, registries, scoring weights, timeouts, cache lifetime, concurrency, or excluded words. CLI flags override file values.
See the configuration guide and scoring methodology.
MCP setup
NameScope exposes the same engine through a local MCP stdio server:
{
"mcpServers": {
"namescope": {
"command": "npx",
"args": ["-y", "namescope", "mcp"]
}
}
}
Available tools are generate_names, check_name, rank_names, check_domains, check_packages, check_github, check_trademark, and explain_score. Read the MCP guide for details.
Development
git clone https://github.com/ensp1re/namescope.git
cd namescope
npm install
npm run check
npm run build
node packages/cli/dist/index.js find "an open-source TypeScript job queue" --offline
The repository is an npm workspace split into schemas, core logic, provider adapters, MCP, and CLI packages. Read the architecture guide for the dependency flow and CONTRIBUTING.md before sending a change.
Know the limits
Results are snapshots, and public namespaces can change at any time. RDAP behavior varies between registries, GitHub search is not exhaustive, local command checks vary by machine, and deterministic quality scores cannot replace research with real people.
Trademark output is preliminary information, not legal advice or a comprehensive clearance search. Consult a qualified trademark attorney before relying on a name commercially. The full list lives in limitations.
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。
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。