API To MCPAI Agent Integration
API To MCPAI Agent Integration

API to MCP Documentation

Build hosted MCP servers for public APIs, SaaS platforms, and internal business systems. Create servers from the visual builder or from your AI agent, then connect them to remote MCP clients such as Codex, Cursor, ChatGPT, Claude, Claude Code, and custom agents.

What You Can Build

  • Convert REST endpoints into hosted MCP tools
  • Connect both REST and GraphQL APIs from one guided builder
  • Expose API tools and workflow tools from one MCP server
  • Use API Key, Bearer Token, Basic Auth, OAuth Client Credentials, or OAuth Authorization Code
  • Let each end user connect their own Google, Meta, CRM, CMS, or internal account through OAuth
  • Use Shopee and TikTok Shop authentication presets, or define a custom OAuth preset in the dashboard or through an agent
  • Shape API responses with standard JMESPath mapping
  • Deploy Open, OAuth/Bearer-protected, or Client Token-protected MCP servers
  • Keep stored secrets encrypted at rest and stripped from forkable snapshots

Overview

MCP (Model Context Protocol) lets AI clients discover and call external tools. API To MCP hosts the MCP runtime for you: you describe the upstream API, define the tools, test them, and deploy a remote HTTP MCP URL.

Core Concepts

Upstream API Authentication

How API To MCP authenticates to the API behind your tool: no auth, API key, Bearer token, Basic Auth, OAuth Client Credentials, or OAuth Authorization Code. Provider presets also describe supported custom OAuth flows and request signing.

MCP Access

How AI clients authenticate before using your MCP server: Open, OAuth/Bearer Token, or Client Token.

Tools

API tools call one endpoint. Workflow tools run multiple API tools in sequence and return one clean result.

Sharing

Listed MCP servers appear in the public directory. Snapshots are forkable revisions that others can fork without using your live server URL.

Security

Stored credentials and OAuth tokens are encrypted at rest, masked in edit screens, and never copied into public snapshots.

Getting Started

You can create MCP servers in two production-ready ways:

Visual Builder

Configure API auth, MCP access, API tools, workflow tools, JMESPath output mapping, testing, and deployment in the dashboard.

AI Agent Builder

Connect https://mcp.apitomcp.app/ and ask an AI agent to create, test, and deploy servers for you.

Quick Example

Want employees to ask an AI agent about internal revenue data?

  1. Add your revenue API: GET /api/revenue
  2. Set authentication, such as OAuth Authorization Code for per-employee access
  3. Map the response with JMESPath so the tool returns only the fields agents need
  4. Deploy and get URL: https://orders.apitomcp.app
  5. Connect the MCP URL to ChatGPT, Claude, Codex, Cursor, or another remote MCP client

Next Steps

Choose your path based on what type of API you're working with:

Quick Start

Simple GET API through the visual builder

Start with basics

Create with AI Agents

Add the manager MCP server to Codex, Cursor, Claude Code, or another remote HTTP MCP client

Connect your agent

MCP Discovery API

List MCP servers and retrieve safe MCP URLs for AI agent platforms and custom clients

View discovery API

Authentication

Understand upstream auth, MCP access, per-user OAuth, and provider presets

Review auth models Use or create an authentication preset

Tools and Workflows

Define API tools, workflow tools, input schemas, and output mapping

Create tools
Next
Quick Start