openapi-to-mcp
Auto-generates MCP tools from your OpenAPI spec, allowing natural language interaction with any API via configurable headers and serverless deployment.
README
openapi-to-mcp
A minimal MCP server that auto-generates tools from your OpenAPI spec. Optimized for Vercel's serverless architecture.
Quick Start
- Fork this repo
- Set
OPENAPI_SPEC_URLto your OpenAPI spec URL - Deploy to Vercel
Configuration
| Variable | Required | Description |
|---|---|---|
OPENAPI_SPEC_URL |
Yes | URL or file path to your OpenAPI spec (YAML or JSON) |
API_BASE_URL |
No | Override the API base URL (defaults to servers[0].url from the spec) |
MCP_SERVER_NAME |
No | Override the MCP server name (defaults to info.title from the spec) |
PORT |
No | Local dev server port (default: 3000) |
How It Works
- On cold start, the server fetches and parses your OpenAPI spec
- Each API endpoint becomes an MCP tool with a Zod-validated input schema
- When a tool is called, the request is forwarded to your API with the caller's headers (API key, Bearer token, etc.) passed through unchanged
- The spec and tools are cached in memory — warm invocations skip the fetch
Running Locally
1. Install dependencies
pnpm install
2. Configure environment variables
Create .env and set OPENAPI_SPEC_URL to your OpenAPI spec. For local dev, this can be a remote URL or an absolute file path:
# Remote URL
OPENAPI_SPEC_URL=https://api.example.com/openapi.yaml
# Or an absolute file path (local dev only)
OPENAPI_SPEC_URL=/path/to/your/openapi.yml
3. Start the dev server
pnpm dev
The server starts at http://localhost:3000/mcp.
4. Connect your MCP client to the local server
How you connect to the MCP server depends on your client (Claude Code, Cursor, VS Code, etc.) — check your client's docs for where to add MCP servers. Most clients use a similar JSON format. Here's an example for Claude Code (.mcp.json):
{
"mcpServers": {
"my-api": {
"type": "http",
"url": "http://localhost:3000/mcp",
"headers": {
"Authorization": "Bearer <your-token>"
}
}
}
}
The headers block above shows a Bearer token, but you can use any header your API requires. The MCP server forwards all headers from your client config to the target API on every tool call without modification — it doesn't validate, store, or care which header is "the auth header". See Auth for examples (API key, Bearer token, multiple headers).
Deploying to Vercel
- Push this repo to GitHub
- Import it at vercel.com/new
- Add
OPENAPI_SPEC_URLunder Settings > Environment Variables.API_BASE_URLandMCP_SERVER_NAMEare optional overrides.
That's it. Your MCP server is live at https://your-project.vercel.app/mcp. Connect clients the same way as local — just swap the URL.
Auth
All headers from your MCP client config are forwarded to the target API on every tool call. The MCP server does not validate or store credentials — your API handles authentication.
Configure whatever headers your API expects in the client config:
// API key auth
"headers": { "X-Api-Key": "<your-key>" }
// Bearer token
"headers": { "Authorization": "Bearer <your-token>" }
// Multiple headers
"headers": { "Authorization": "Bearer <token>", "X-Org-Id": "org_123" }
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。