customclaw-mcp
Registry of 37 vetted single-file utilities for LLM and agent code (rate limiters that respect Retry-After, JSON repair, response caches, retry with backoff, token counters, PII scrubbers). 4 MCP tools: list_utilities, search_utilities, get_utility_info, install_utility.
README
customclaw-mcp
Install, don't regenerate. MCP server for the CustomClaw registry — 37 vetted single-file utilities for LLM and agent code (rate limiters that respect
Retry-After, JSON repair, response caches, retry with backoff, token counters, PII scrubbers). Your agent gets four native tool calls instead of 120 lines of plausible-looking new code.
📖 Background: The utility-amnesia problem — why your coding agent rewrote the same rate limiter 47 times.
30-second install
npx -y customclaw-mcp
Drop that command into your MCP client's config (snippets below) and you're done. Free tools install with no auth.
What it does
Plugs the CustomClaw catalog (37 utilities, 32 free) into any Model Context Protocol host — Claude Desktop, Claude Code, Cursor, Windsurf, anything that speaks MCP — as four tools:
| Tool | Arguments | Purpose |
|---|---|---|
list_utilities |
— | Full catalog (cached 5 min) |
search_utilities |
query: string |
Fuzzy match on slug/name/description |
get_utility_info |
slug: string |
Full detail for one utility |
install_utility |
slug, target_dir, session_id? |
Fetch payload + write files |
Install (by client)
Claude Desktop
~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"customclaw": {
"command": "npx",
"args": ["-y", "customclaw-mcp"]
}
}
}
Restart Claude Desktop. Tools appear in the tool picker.
Claude Code
claude mcp add customclaw -- npx -y customclaw-mcp
Cursor / Windsurf / any stdio MCP host
Point the host at npx -y customclaw-mcp.
Streamable HTTP (smithery, ChatGPT, etc.)
Use the hosted endpoint: https://customclaw.company/mcp. Registered on smithery as onlythebestswfl/customclaw.
Example prompts
Once connected:
- "Add a rate limiter that respects Retry-After." →
search_utilities→install_utility rate-limit-handler - "I need to parse broken JSON from Claude output." →
json-repair - "What do you have for caching LLM responses?" →
search_utilities→ short results table - "Install
token-optimiserinto./src/lib." → directinstall_utility - "Scan this project's prompts for injection — is there a tool?" →
injection-scanner
Task → slug quick reference
| If you need… | Slug |
|---|---|
Rate limiting with Retry-After handling |
rate-limit-handler |
| JSON repair for LLM output | json-repair |
| Response cache by prompt hash | response-cache |
| Retry with exponential backoff | retry-with-backoff |
| Token counting across models | token-optimiser |
| PII scrubbing before logging | pii-scrubber |
| Prompt-injection scanner | injection-scanner |
| Cost forecaster per provider | cost-forecaster |
Full list via list_utilities or customclaw.company/api/catalog.
Paid utilities
For paid utilities, pass the Stripe checkout session_id from your receipt email as the session_id argument to install_utility. Without it, the tool returns a helpful 401.
Environment
CUSTOMCLAW_BASE— registry base URL override (defaulthttps://customclaw.company). Useful for staging or self-hosted mirrors.
How it works
- Node.js stdio MCP server. Dependency:
@modelcontextprotocol/sdk. - Catalog fetched from
GET /api/catalog; cached 5 min per process. - Payloads from
GET /api/cli?slug=<slug>[&session_id=<cs_...>]— same endpoint the officialcustomclaw-cliuses. install_utilityrefuses to write outsidetarget_dir(path-traversal guard).
Links
- Registry: customclaw.company
- CLI companion:
customclaw-cli - Source: github.com/onlythebestswfl-ops/customclaw-mcp
License
MIT.
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。