Skip to content
N
Booting Engineering OS
000%
All projects
Active#3 · Wedge (build first)

MCP Server Generator

Turn any API into a production-ready, well-described MCP server in minutes — secure-by-default, not a toy.

The fastest way to make any API callable by AI — production-ready MCP servers, generated.

View repoMCP tooling
Status
Active build (MCPForge). Generation engine + CLI + Next.js web app with Stripe wiring. Apache-2.0.
Difficulty
Medium
Time to MVP
30 days (solo-dev MVP)

Role in the platform

Top-of-funnel and the MCP teacher. Every generated server is a lead for #2 and #1; it teaches the MCP layer the whole family depends on.

Overview

A developer with an API wants an AI agent (Claude, Cursor) to use it via MCP. Doing that by hand is a half-day to days of boilerplate — transport, auth, validation, and the hard part: tool descriptions good enough that the model calls tools correctly.

MCPForge automates it: paste an OpenAPI spec, pick endpoints, and download a secure-by-default, well-described, runnable MCP server in under three minutes.

The problem

Official SDKs and boilerplate require manual wiring; OpenAPI→tool converters are generic, not MCP-idiomatic, and ship no auth, hosting, or observability.

Most generators produce toy servers, not safe ones — and they skip the tool descriptions, which is exactly the part LLMs need most.

What it is

A generation engine + CLI (zero-dependency TypeScript) and a Next.js web app: paste a spec, preview, and download a ZIP with a runnable TypeScript MCP server, auth scaffolding, Zod validation, README, and mcp.json.

Free to generate; Pro unlocks unlimited endpoints, premium tool descriptions, GitHub push, and history; a done-for-you setup is the fastest path to first revenue.

Key features

OpenAPI → MCP server

Paste a spec, pick endpoints, download a runnable TypeScript MCP server in under three minutes.

Great tool descriptions

LLM-generated tool descriptions tuned so the model calls tools correctly — the hardest, highest-value part.

Secure by default

Auth scaffolding, Zod input validation, and sane defaults baked into every generated server.

Runs where agents run

Generated servers run in Claude Desktop / Cursor; mcp.json + README included.

Pro: push & history

Unlimited endpoints, premium descriptions, one-click GitHub push, and generation history.

Tech stack

Next.js 14TypeScriptTailwind + shadcn/ui@modelcontextprotocol/sdk@apidevtools/swagger-parserAnthropic SDKZod + jszipPostgres (Neon) + DrizzleClerk · Stripe · Octokit

Why it's different

  • Production-ready, secure-by-default output — not toy servers.
  • Generates the tool descriptions that determine whether an LLM uses the API correctly.
  • Cheapest to ship, fastest to value, most shareable product in the family — and a lead source for the rest.

Roadmap

MVP

OpenAPI → TS MCP server, LLM tool descriptions, auth scaffold, Zod validation, preview + ZIP, Free/Pro + Stripe, GitHub push.

Next

Postgres/GraphQL/codebase inputs, hosting/runtime, premium description tiers.

Later

Teams, observability, sandboxes, Python output.

Platform

Feeds servers into ContextOS (#1) and the Agent Marketplace (#7).