API To MCP

API To MCP is a hosted platform that converts REST and GraphQL APIs into Model Context Protocol (MCP) servers, deployable in minutes from a visual dashboard or directly from an AI coding agent. It targets developers and teams who want to expose internal business systems, SaaS platforms, or public data APIs as tools that AI agents like Codex, Cursor, Claude Code, and ChatGPT can call over remote HTTP.

What It Is

API To MCP sits between an upstream API and an AI agent: you describe the API, configure authentication, define tools, and the platform hosts a Streamable HTTP MCP endpoint at a dedicated apitomcp.io URL. The platform handles the MCP runtime, SSL, credential encryption, and usage tracking so teams don't need to write or maintain custom MCP server code. It supports both REST and GraphQL APIs and covers a wide range of upstream authentication models — no auth, API key, Bearer token, Basic Auth, OAuth Client Credentials, and per-user OAuth Authorization Code.

Two Build Paths

The platform offers two production-ready creation flows:

• Visual Builder — A guided dashboard where teams configure the base URL, upstream auth, MCP access policy, API tools, workflow tools, JMESPath response mapping, and run endpoint tests before deploying.
• Agent Builder — Connect the API To MCP manager server (https://mcp.apitomcp.io/) to Codex, Cursor, Claude Code, or another remote MCP client, then describe the API in chat and let the agent create, update, test, and deploy the server.

Both paths produce the same hosted MCP endpoint; the choice depends on whether the team prefers dashboard control or agent-driven iteration.

Authentication Architecture

API To MCP separates upstream API authentication from MCP server access. Upstream auth covers how the platform calls the target API; MCP access covers how AI clients connect to the MCP server itself. Stored credentials — API keys, Bearer tokens, Basic passwords, OAuth client secrets, access tokens, and refresh tokens — are encrypted at rest and masked in owner-facing edit screens. Snapshots used for sharing never include live secrets or active connection tokens. The about page notes that production encryption keys can be isolated from the application database through AWS Secrets Manager and KMS-backed key handling.

Tool Composition and Output Shaping

Beyond simple one-endpoint API tools, the platform supports workflow tools that chain multiple API tool calls into a single MCP tool, enabling multi-step lookup, enrichment, and action flows. Responses can be shaped with JMESPath mapping to return clean, structured JSON that is easier for AI agents to consume, with a generated output schema attached to each tool.

Compatibility and Deployment

Deployed MCP servers are compatible with ChatGPT, Claude, Codex, Cursor, Claude Code, Visual Studio Code, and custom agents that support remote HTTP MCP. Each server gets a hosted Streamable HTTP endpoint with SSL, open or authenticated access modes (Open, OAuth/Bearer Token, or Client Token), and usage tracking. Teams can also publish servers to a public MCP directory or share forkable snapshots that copy tools and configuration without secrets.

Current Status

The platform is actively available with a free tier requiring no credit card. The pricing page notes launch promotional pricing across paid tiers, signaling the product is in an early commercial phase. The about page describes the team as experienced in API design, AI systems, and developer tooling, building toward making AI agent integration accessible without custom integration code.