mcp-swagger
MCP (Model Context Protocol) server for exposing Swagger/OpenAPI documentation to AI models like Claude. This tool allows AI assistants to explore, understand, and interact with your API documentation seamlessly.
claude mcp add --transport stdio awssam-mcp-swagger npx @awssam/mcp-swagger https://petstore.swagger.io/v2/swagger.json
How to use
The mcp-swagger server exposes Swagger/OpenAPI documentation to AI models by providing a structured, queryable interface for exploring endpoints, schemas, tags, and API metadata. It includes 14 tools designed for API discovery, endpoint details, curl command generation, schema analysis, and even safe execution against a live API. This lets models like Claude understand and reason about an API without needing to interpret raw documentation, enabling tasks such as curl generation, endpoint lookups, and impact analysis directly from the OpenAPI spec. To use it, run the MCP server (for example via npx) and configure your MCP consumer to point at the provided server to access the available tools and data.
How to install
Prerequisites:
- Node.js (v18+) and npm installed on your system
- Internet access to fetch the MCP package
Install the MCP Swagger server globally:
npm install -g @awssam/mcp-swagger
Run the server locally (example using the Petstore OpenAPI document):
npx @awssam/mcp-swagger https://petstore.swagger.io/v2/swagger.json
Alternatively, run via Claude CLI configuration as shown in the Quick Start:
# Claude CLI (Recommended) - add via CLI or config
claude mcp add swagger npx @awssam/mcp-swagger https://petstore.swagger.io/v2/swagger.json
Or place the configuration in Claude's YAML config and restart Claude:
mcpServers:
swagger:
command: npx
args:
- "@awssam/mcp-swagger"
- "https://petstore.swagger.io/v2/swagger.json"
If you have a local API URL, you can export it and run:
export API_URL=https://your-api.com/api-docs
npx @awssam/mcp-swagger
Additional notes
Tips and notes:
- The server supports OpenAPI 3.0 and is compatible with Swagger 2.0 endpoints.
- It caches fetched Swagger documents automatically for faster responses.
- Use the executeEndpoint tool to test endpoints and fetch live responses when needed.
- Ensure your OpenAPI URL is publicly reachable or accessible from where the MCP server runs.
- If you customize environment variables, document them in your deployment config (e.g., API_URL, CACHE_TTL).
- If endpoints change, reload the MCP server or clear the cache to fetch the latest OpenAPI document.
- Useful for integration with Claude’s mcp workflow via the 14 available tools (listEndpoints, getEndpointDetails, searchEndpoints, etc.).
Related MCP Servers
mongodb-lens
🍃🔎 MongoDB Lens: Full Featured MCP Server for MongoDB Databases
mcp-arr
MCP server for *arr media management suite
unity-api
Instant, accurate Unity API lookups instead of expensive source file reads, saving your agent tokens, context, and hallucinations
unreal-api
Instant, accurate Unreal Engine API lookups instead of expensive source file reads, saving your agent tokens, context, and hallucinations.
short-url
简单易用的短链接生成工具,完全开源、免费、无需登录,可私有化部署,链接永久有效!
skill-to
Convert AI Skills (Claude Skills format) to MCP server resources - Part of BioContextAI