ToolHail
Discovers MCP servers by capability and lets agents call them remotely through a hosted proxy, with per-use metering and payment via Stripe.
README
ToolHail
The discovery-and-payment layer for AI agents. ToolHail lets an AI agent find the right MCP (Model Context Protocol) server for any task and call it — paying per use — without installing anything.
There are 10,000+ MCP servers an agent could use, and no good way to discover and use one mid-task. ToolHail closes that gap with two tools:
- 🔎
find_mcp_server(free) — describe a capability in plain language ("query a Postgres database","send a Slack message","scrape a web page") and get ranked, deduplicated MCP servers from every major registry (mcp.so, PulseMCP, Smithery, Glama) with ready-to-paste setup. - ⚡
call_tool(paid) — invoke a discovered remote MCP server through ToolHail's SSRF-guarded proxy. No install, one hop; each successful call is metered and billed through Stripe.
Use it
Hosted MCP endpoint (Streamable HTTP):
https://mcp.toolhail.com/mcp
Add it to Claude Code (or any MCP client):
claude mcp add --transport http toolhail https://mcp.toolhail.com/mcp
MCP Scout on the Apify Store — the discovery engine as a standalone, pay-per-search actor: https://apify.com/01010101/mcp-scout
Site: https://toolhail.com · Guide: https://toolhail.com/guide
How it works
- Discover — the agent hits a capability gap and asks
find_mcp_server, which searches the registries and returns the best-fit servers, ranked for usability. - Execute — instead of installing, the agent calls the chosen tool via
call_tool; ToolHail proxies it in one hop, nothing to configure. - Settle — each successful call is metered and billed through Stripe. The agent pays for exactly what it used.
Why it exists
An agent's toolset is fixed at configuration time, but real tasks need capabilities the agent doesn't have. Today the options are "hard-code every server you might ever need" or "give up." ToolHail makes tools a runtime resource: discovered on demand, called through one endpoint, paid per use. It is discovery and payment infrastructure for the agentic web — built entirely on the open Model Context Protocol.
Repo layout
| Path | What |
|---|---|
apify-mcp-scout/ |
MCP Scout — the discovery engine, published on the Apify Store (live) |
gateway-worker/ |
Hosted MCP gateway on Cloudflare Workers — find_mcp_server + call_tool (live at mcp.toolhail.com) |
gateway/ |
Reference gateway + Stripe metering + tests |
landing/ |
toolhail.com site (landing + guide + llms.txt) |
registry/ |
MCP registry listing metadata (server.json) + submission playbook |
dashboard/ |
Private revenue/usage dashboard |
brand/ |
Brand assets |
Built on
Model Context Protocol · Cloudflare Workers · Stripe · Apify
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。