swagger-mcp
A read-only MCP server for navigating OpenAPI / Swagger specifications, enabling agents to search endpoints, retrieve parameters and schemas, and inspect authentication without loading the full spec into context.
README
swagger-mcp
A read-only Model Context Protocol server for navigating an OpenAPI / Swagger specification. It gives a coding agent tools to explore a large API's documentation — search endpoints, read parameters and schemas, inspect auth — without dumping the whole spec into context.
It never calls the described API. It only reads its documentation.
Works with any MCP-capable harness (Claude Code, Cursor, Cline, Windsurf, Zed, opencode, …) over stdio.
Usage
Add it to your MCP client config. No install step — npx fetches it on demand:
{
"mcpServers": {
"swagger-mcp": {
"command": "npx",
"args": ["-y", "@emreyc/swagger-mcp@latest"]
}
}
}
Every tool takes a spec parameter — a URL (or local path) to a Swagger 2.0 or
OpenAPI 3.x document — so a single server can navigate any number of specs in one
session. Internal/localhost URLs (e.g. Spring Boot's
http://localhost:8080/v3/api-docs) are supported.
If you mostly work with one spec, set API_DOCS_URL as a default so spec can be
omitted:
{
"mcpServers": {
"swagger-mcp": {
"command": "npx",
"args": ["-y", "@emreyc/swagger-mcp@latest"],
"env": { "API_DOCS_URL": "https://api.example.com/swagger.json" }
}
}
}
Specs are fetched lazily and cached per URL for the process lifetime.
Tools
Designed for progressive disclosure — orient, then discover, then drill down — so the agent loads only what it needs.
| Tool | Purpose |
|---|---|
get_api_overview |
Title, version, servers, global security, counts. Start here. |
list_tags |
Tags (logical groups) with endpoint counts. |
list_endpoints |
Compact index of endpoints; optional tag filter. |
search_endpoints |
Substring search over path/summary/description/params. |
list_schemas |
Named schemas with one-line descriptions. |
search_schemas |
Substring search over schema names, fields, descriptions. |
get_endpoint |
Full detail for one method+path: params, request/response schemas, auth. |
get_schema |
Resolve one named schema. |
get_auth |
Security schemes in full (OAuth2 flows, scopes, API-key locations). |
Every tool also accepts an optional spec argument (a spec URL or path); when
omitted it falls back to API_DOCS_URL.
Swagger 2.0 and OpenAPI 3.x are normalized into one consistent output shape.
Nested named schemas are shown as { "$schema": "Name" } markers — call
get_schema to expand them. This keeps every response bounded and is safe with
circular schemas.
Development
npm install
npm run typecheck
npm test
npm run build
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。
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。
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。
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。