Quickstart

How onboarding works

MCPify is a hosted, multi-tenant gateway. You do not install or run anything: we configure your API as a service on the gateway, and you point your AI client at the resulting MCP endpoint. There is no self-serve console and no CLI — onboarding starts with a conversation, because the first question is always which API you are exposing and what the agent should be allowed to do with it.

What we need from you

Step 1: Tell us about the API

Start with the contact form. We will come back with what the generated tool set would look like for your API — the tool names, their descriptions, and the shape of what the agent gets back.

Step 2: We configure the service

We register your API as a service on the gateway: importing its schema, generating tool descriptions the model can actually act on, and configuring pagination, caching, and rate limits for its behaviour. Once an integration is agreed, you submit the upstream credentials directly through credentials.mcpify.org — they are stored in Google Secret Manager and never exposed to the AI client.

What the gateway does with those tool descriptions is the interesting part; see Tool Descriptions and MCP Schema.

Step 3: Connect your AI client

Your service gets its own endpoint on the gateway, speaking MCP over streamable HTTP:

my-api is your service name, assigned when we configure it. Each service has its own subdomain, which is what keeps OAuth state and cookie scope isolated between services.

from anthropic import Anthropic

client = Anthropic()
response = client.messages.create(
    model="claude-sonnet-5",
    mcp_servers=[{
        "url": "https://my-api.mcp.mcpify.org/mcp"
    }],
    messages=[{"role": "user", "content": "List all users"}]
)

from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-5",
    tools=[{
        "type": "mcp",
        "server_url": "https://my-api.mcp.mcpify.org/mcp"
    }],
    input="Get the list of active users"
)

Next Steps