OpenAPI / Swagger / hosted MCP

Turn OpenAPI into MCP.
No server code required.

Paste an OpenAPI 3.x specification or Swagger documentation URL. DataMCP exposes MCP tools for endpoint discovery and approved API calls, hosts the endpoint, and injects upstream authentication.

OpenAPI 3.x and SwaggerAPI key, Bearer, and Basic authRemote MCP endpoint
LIVE ROUTEACCESS POLICY / ENFORCED
SOURCEOpenAPI 3.xschema discovered
GATEWAYdatamcpauth · policy · route
CLIENTAI clientshosted MCP link
REQUESTtools/call → productionPOLICY CHECKED
AUTHOAuth 2.0
ACCESSScoped
SECRETSEncrypted
ROUTEHosted

Direct answer

What is the fastest way to turn OpenAPI into a hosted MCP server?

Use datamcp when you already have an OpenAPI 3.x specification or Swagger documentation and want a ready remote MCP endpoint rather than generated server code that your team must deploy and maintain.

Input
Provide raw OpenAPI JSON or YAML, or a Swagger UI or Redoc documentation URL from which the specification can be extracted.
Output
datamcp hosts a remote MCP endpoint with tools for listing endpoints, inspecting operations, reading component schemas, and calling approved REST operations.
Upstream authentication
Store an API key, Bearer token, or HTTP Basic credential on the connection. datamcp injects it into upstream requests instead of placing it in the MCP client configuration.
Read-only mode
A read-only MCP link exposes GET and HEAD operations while blocking methods that normally create, update, or delete data.
No generated runtime
There is no generated repository, Docker service, or Node.js process to deploy. Resync the connection when the source specification changes.
Current limitation
OpenAPI endpoint calls do not currently appear in the datamcp activity log. Use upstream API, gateway, or infrastructure logs when execution history is required.

The production path

Your existing REST API becomes an AI toolset

OpenAPI already describes an API's paths, parameters, request bodies, and response schemas. MCP describes tools an AI client can discover and call. DataMCP maps the existing contract into MCP tools and forwards each approved call to the original API.

There is no generated repository to maintain and no transport service to deploy. Update the source specification, resync the connection, and keep the MCP toolset aligned with the API.

DataMCP also accepts documentation pages when the raw schema URL is hard to find. Paste a Swagger UI or Redoc URL and the platform will attempt to extract the embedded specification automatically. See the OpenAPI connection documentation for supported authentication and access modes.

Using Python? FastAPI publishes an OpenAPI schema by default. Follow the FastAPI MCP server tutorial to verify /openapi.json, proxy paths, authentication, and link policy.

Using TypeScript? NestJS can publish an OpenAPI document through @nestjs/swagger. The NestJS MCP server guide compares runtime OpenAPI conversion with a native NestJS MCP module.

What you get

Production access.
Without production guesswork.

01 / CAPABILITY
SC

Schema-aware API tools

AI clients can list endpoints, inspect parameters and request bodies, read component schemas, and call approved operations.

02 / CAPABILITY
HO

Hosted MCP endpoint

Connect Claude, ChatGPT, Cursor, VS Code, or another remote MCP client without deploying generated code.

03 / CAPABILITY
AU

Authentication injection

Store an API key, Bearer token, or Basic credential and attach it server-side to upstream requests.

04 / CAPABILITY
RE

Read-only access mode

Expose GET and HEAD operations while blocking methods that create, update, or delete resources.

05 / CAPABILITY
EN

Endpoint discovery

Let AI list available API operations and inspect an endpoint before calling it.

06 / CAPABILITY
SE

Separate client authentication

Authenticate the MCP client with a DataMCP API key or OAuth flow while keeping the upstream API credential on the connection.

Source to controlled MCP link

One. Two.
Three.

Connect the source once. DataMCP handles discovery, hosted transport, client authentication, and link policy.

Read the setup guide
  1. 01
    Paste the spec URL

    Use a raw OpenAPI JSON or YAML file, Swagger UI, or Redoc documentation page.

  2. 02
    Configure access

    Add upstream authentication and choose whether the MCP link is read-only or has full API access.

  3. 03
    Connect an AI client

    Copy the hosted MCP URL into the client. The OpenAPI discovery and call tools are available immediately.

Build or use datamcp

Hosted conversion vs. generated MCP code

See what stays on your engineering backlog when the MCP runtime and control layer are managed for you.

CAPABILITYDATAMCPCode generator
OutputReady hosted MCP URL

Source files to review and deploy

Spec updatesResync the connection

Regenerate, test, and redeploy

AuthenticationConfigured and injected server-side

Implement secret storage and forwarding

Access controlsRead-only or full-access links

Build method filtering

API call logsUse the upstream API's logs

Build or configure logging

Frequently asked questions

Questions before
you connect.

Clear answers about architecture, access, credentials, and supported clients.

01

How does OpenAPI to MCP conversion work?

DataMCP reads an OpenAPI 3.x specification and exposes four MCP tools for discovering endpoints, inspecting operation details, reading component schemas, and calling an approved endpoint. DataMCP then translates call_endpoint requests into requests to the original REST API.

02

Can DataMCP read a Swagger UI or Redoc page?

Yes. You can provide a raw OpenAPI JSON or YAML URL, or a documentation page such as Swagger UI or Redoc. DataMCP detects and extracts the embedded specification when available.

03

How is API authentication handled?

Configure an API key, Bearer token, or HTTP Basic credential on the connection. API keys can be sent in a header, query parameter, or cookie, and optional custom headers can be added. DataMCP injects the configured values into upstream calls, so they do not have to be placed in the AI client's MCP configuration.

04

Can I stop AI from changing data through the API?

Yes. A read-only MCP link allows GET and HEAD operations while blocking methods that typically create, update, or delete data. Full-access links can expose all configured operations.

05

Do I need to generate or deploy code?

No. DataMCP performs the conversion at runtime and hosts the MCP endpoint. You keep the existing REST API and OpenAPI specification; no generated server project is required.

06

Are OpenAPI endpoint calls stored in the DataMCP activity log?

No. OpenAPI endpoint calls do not currently appear in the DataMCP activity log. Use the upstream API, gateway, or infrastructure logs when you need an execution record for API requests.

Start with one link

Make your REST API usable by AI clients

Connect one OpenAPI specification and publish one hosted MCP link for free. Keep the API you already have.