The standard format for defining REST APIs. AI agents cannot read code, but they can parse an OpenAPI schema to discover your backend. With wmcp.sh, you can turn any OpenAPI spec directly into native Model Context Protocol (MCP) tools.
If you have an existing REST API built in Node, Python, PHP, or Java, you likely already generate an OpenAPI spec for human documentation (like Swagger UI). But modern AI frameworks like LangChain, OpenAI, and Anthropic's Claude Desktop require a specific "tool_use" or "function calling" JSON schema format to interact with external systems.
Manually rewriting your OpenAPI paths, request bodies, and security schemes into agent-compatible tools is a massive maintenance burden. Instead of rewriting, you can bridge the gap dynamically.
# Pass any valid OpenAPI v3 URL to wmcp.sh's tool extractor API
curl -X GET "https://wmcp.sh/api/v1/tools?url=https://api.example.com/openapi.json" \
-H "Authorization: Bearer YOUR_WMCP_KEY"
# wmcp.sh returns an array of MCP-compatible tool definitions
# Example output:
[
{
"name": "createOrder",
"description": "Create a new order in the system",
"inputSchema": {
"type": "object",
"properties": {
"sku": { "type": "string" },
"quantity": { "type": "integer" }
},
"required": ["sku", "quantity"]
}
}
]| Capability | Without wmcp.sh | With wmcp.sh |
|---|---|---|
| Tool generation | ⚠️ Hand-written JSON schemas | ✅ Extracted directly from OpenAPI paths |
| Schema sync | ❌ Manual updates required when API changes | ✅ Always matches your live OpenAPI spec |
| Security & Auth | ⚠️ Hardcoded tokens in agent logic | ✅ Auto-injected via OpenAPI securitySchemes |
| Framework support | ⚠️ Write adapters for LangChain, OpenAI, etc. | ✅ Native MCP protocol, compatible everywhere |
| Route filtering | ❌ All or nothing exposure | ✅ Filter by OpenAPI tags (e.g. ?tag=agent) |
Swagger was the original name for the specification (up to version 2.0). It was donated to the Linux Foundation and renamed OpenAPI Specification (OAS) from version 3.0 onwards. Swagger now refers to the tooling ecosystem (like Swagger UI).
AI agents from Anthropic and OpenAI cannot read your backend source code. An OpenAPI spec provides a machine-readable map of your endpoints, inputs, and outputs, which can be translated into function calling schemas.
wmcp.sh primarily supports OpenAPI v3.0 and v3.1, which are the modern standards. v2.0 (Swagger) specs are usually auto-upconverted or can be easily migrated.
No, MCP is protocol-agnostic. However, if you already have a REST API, providing an OpenAPI spec is the fastest way to generate an MCP server, as platforms like wmcp.sh can convert it dynamically.
Custom adapter + hosted MCP at mcp.yourbrand.com + verified badge. From $499 one-time setup.