Model Context Protocol Platform

Turn any REST API into an MCP server

API2MCP wraps your existing APIs as Model Context Protocol servers — so Claude, Cursor, and other AI agents can call them directly. Auto-discover endpoints from OpenAPI specs, manage credentials securely, and connect AI clients in minutes.

Get Started Free Read the Docs

Free tier available · No credit card required

Everything you need to bridge APIs and AI

Six core capabilities, one unified platform. Built for teams who want to expose their APIs to AI agents without writing custom MCP server code.

API Provider Management

Register any REST API with its base URL, OpenAPI spec, and authentication. Add as many as your tier allows.

One-Click MCP Servers

Wrap any registered provider as an MCP-compatible server with a unique URL and access token — no code required.

Automatic Tool Discovery

API2MCP parses your OpenAPI/Swagger spec and turns each endpoint into a structured MCP tool the AI can call.

Encrypted Credentials

API keys, bearer tokens, and OAuth secrets are encrypted at rest and injected server-side. AI clients never see your secrets.

Multiple Auth Types

Bearer tokens, API keys, and OAuth 2.1 — configure once and we handle the auth flow on every upstream request.

Usage Monitoring

Live dashboard of providers, servers, tools, and request volume. Stay inside your tier limits with real-time visibility.

See API2MCP in action

Short walkthroughs of the real dashboard, from pasting a docs URL to an AI agent calling your API.

Import from a URL

Point API2MCP at an API's documentation page and it finds the spec, the base URL, and the auth type on its own. OpenAPI, Swagger, WSDL, GraphQL, Postman, and HAR all work, on the free tier.

  1. You paste a docs URL and click Detect API.
  2. API2MCP tries the URL, the docs page, then the standard spec paths.
  3. It finds the spec, the base URL, and the auth type.
  4. One click imports the provider, ready to wrap.

Find it with AI

When an API has no machine-readable spec, AI reads the docs and drafts an OpenAPI definition of its most useful endpoints, marked unverified until a real call succeeds. Once one does, the spec joins the shared catalog and the next person importing that API gets it instantly, with no AI call.

  1. Detection found the API, but no endpoints.
  2. AI searches the web and drafts an OpenAPI definition.
  3. The draft arrives marked unverified, with confidence and source.
  4. The first successful call promotes the spec to the shared catalog.

Wrap and connect

Wrap any provider as an MCP server with its own URL and access token, then drop the generated config into Claude Code, Cursor, or Windsurf. Two servers are included on the free tier.

  1. You click Wrap to MCP and pick how clients authenticate.
  2. API2MCP creates the server and shows the token once.
  3. You paste the generated config into Claude Code and connect.
  4. The server goes Live and counts every call.

Tools, already curated

Each endpoint becomes an MCP tool with a typed input schema, and discovery cleans up as it goes: duplicate API versions collapse, bulk and file endpoints start switched off, and your API key never appears in a tool's inputs. Fewer, better tools mean the model picks the right one.

  1. Refresh Tools reads the spec and makes one tool per endpoint.
  2. Duplicate API versions collapse into one tool; the highest wins.
  3. Bulk and file endpoints start off and sort last.
  4. Validate calls the live API and spots a wrong base URL.
  5. One click applies the fix; every tool comes back init ok.

Suggest with AI

Coming soon

Ask AI to review your tool list and it proposes clearer names, agent-oriented descriptions, and which tools to hide. You review every change side by side and adopt only what you want; nothing changes until you say so.

  1. You click Suggest with AI; the analysis runs in the background.
  2. A compare view shows Original beside Proposed, change by change.
  3. You tick the changes you want and adopt them.
  4. Adopted tools wear an AI badge and survive Refresh Tools.

Secrets stay server-side

Two auth layers: clients authenticate to your MCP server, and API2MCP holds the upstream credentials, encrypted at rest and injected at call time. Every outbound call is checked against private addresses, and a rate-limited upstream returns a retry hint instead of an error.

  1. Claude calls a tool with only the server's bearer token.
  2. API2MCP checks the token, then injects your API key from the vault.
  3. The API responds; the result flows back, the key stays put.
  4. A rate-limited upstream returns a retry hint, not a failure.

From API to AI tool in five steps

No code, no glue scripts, no custom MCP server boilerplate.

  1. 1
    Register a provider
    Add your API's base URL, authentication, and OpenAPI spec.
  2. 2
    Create an MCP server
    Get a unique server URL and access token in seconds.
  3. 3
    Auto-discover tools
    Each endpoint becomes a structured MCP tool automatically.
  4. 4
    Connect your AI client
    Point Claude Desktop, Cursor, or any MCP client at your server.
  5. 5
    AI calls your API
    Credentials stay server-side. The AI sees tools and responses, never secrets.

Plans that scale with you

Start free. Upgrade when your AI agents start doing real work.

  • Free — $0/month: 2 MCP server connections, 1,000 requests/day, 10 req/min, 1 MB max payload, community support.
  • Pro — $39/month: 10 MCP servers, 25,000 requests/day, 100 req/min, 10 MB payload, AI Spec Discovery, full 30-day dashboard.
  • Business — $149/month: Unlimited MCP servers, 100,000 requests/day, 500 req/min, 25 MB payload, AI Spec Discovery.
  • Enterprise — custom: SSO/SAML, dedicated infrastructure, unlimited limits. Annual billing saves up to 17%.

Frequently asked questions

What is the Model Context Protocol (MCP)?

The Model Context Protocol (MCP) is an open standard that lets AI agents such as Claude and Cursor call external tools and APIs in a structured way. An MCP server exposes named tools; an MCP client discovers and invokes them. API2MCP implements the server side automatically from any OpenAPI or Swagger spec.

What is API2MCP?

API2MCP is a SaaS platform that wraps any REST API as an MCP server. You register your API's base URL, authentication, and OpenAPI spec; API2MCP turns each endpoint into a structured MCP tool and issues a server URL and access token. Claude Desktop, Cursor, and other MCP clients can then call your API — no MCP server code required.

How does API2MCP work?

Five steps: register a provider with its base URL, authentication, and OpenAPI spec; create an MCP server and get a unique URL and access token; endpoints auto-discover into MCP tools; point Claude Desktop, Cursor, or any MCP client at the server; the AI calls your API while credentials stay server-side.

Is API2MCP secure?

API keys, bearer tokens, and OAuth secrets are encrypted at rest and injected server-side on every upstream request. AI clients receive only the MCP tool responses, never the underlying credentials. Bearer token, API key, and OAuth 2.1 authentication flows are supported.

Which AI clients work with API2MCP?

Any MCP-compatible client works out of the box, including Claude Desktop, Cursor, and any agent implementing the MCP client specification. API2MCP issues a standard MCP server URL and access token you add to your client configuration.