Manufact - Build and Deploy MCP Servers and AI Agents

TL;DR

TL;DR: Manufact is a YC S25-backed platform combining the open-source mcp-use SDK with a managed cloud for building, testing, and deploying MCP servers and AI agents.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: manufact.com
• Source repository: github.com/mcp-use/mcp-use — verified MIT license, README confirms SDK description
• License: MIT (verified via GitHub API license.spdx_id)
• HN launch thread: news.ycombinator.com/item?id=42725306 — 111 points, YC S25 batch
• Source last checked: 2026-08-18 (commit verified from GitHub API)

What Is Manufact?

Manufact is a developer platform for building and deploying MCP (Model Context Protocol) servers and AI agents. It has two components:

mcp-use — the open-source SDK (MIT licensed, 10.5k GitHub stars as of August 2026) for building MCP servers, ChatGPT Apps, and Claude connectors in TypeScript and Python.

Manufact Cloud — the hosted deployment platform built on top of mcp-use, adding preview environments, cross-client testing, publishing checks, and usage analytics.

From the README:

The TypeScript framework for MCP. Build, test, and ship MCP servers, ChatGPT plugins, Claude connectors.

Key SDK features from the README:

• Fully typed — Zod schemas flow from tools to structured results, View props, and tool calls
• Native Views — bind React Views directly to tools and ship interactive apps without custom extension wiring
• Agent-first and headless — scaffold, invoke, inspect, screenshot, and deploy through your agent
• Built-in Inspector — inspect tools and Views in the browser or headlessly via CLI

Setup Workflow

Prerequisites

• Node.js 18+ or Python 3.10+
• A Manufact Cloud account (free tier works for local development)

Step 1: Install the SDK

# TypeScript
npm install mcp-use

# Python
pip install mcp-use

Step 2: Scaffold a project

npx -y create-mcp-use-app@latest
cd your-project
npm install

Step 3: Run locally

npm run dev
# Opens http://localhost:3000/mcp/inspector

Step 4: Deploy to Manufact Cloud

Connect your GitHub repository from the Manufact dashboard, or use the CLI:

# Connect an existing MCP server by URL (no hosting required)
# Useful for testing against a running server before deploying

The Free plan at $0/month includes $5 in usage credits. Paid plans start at $25/month (Hobby) with preview deployments and E2E checks.

How Deployment Works

Manufact Cloud deploys MCP apps from GitHub with the following flow:

1. Connect repo — authorize GitHub access from the Manufact dashboard
2. Preview per branch — every branch gets its own MCP URL (<slug>--<branch>.run.mcp-use.com/mcp) for pre-merge testing with real AI clients
3. Publishing checks — audit your MCP app against ChatGPT Apps Store and Cloud Connectors requirements before going live
4. Cross-client testing — run the same test suite against ChatGPT, Claude Desktop, and other MCP-compatible clients simultaneously
5. Usage analytics — per-tool latency, error rates, and token usage across all clients

Pricing

From manufact.com/pricing:

| Plan | Price | Notes | |------|-------|-------| | Free | $0/mo | $5/mo usage credits, MCP URL for local dev | | Hobby | $25/mo | Preview deployments, E2E checks, submission pack | | Startup | $250/mo | All regions | | Enterprise | Custom | |

Usage is metered:

• $0.10 per 1,000 tool-call requests
• $1 per eval run
• $2 per E2E check

Discovery traffic (tools/list, etc.) is never billed.

Open-Source vs. Cloud

You can use mcp-use without Manufact Cloud entirely:

• Local development — npm run dev with the built-in Inspector (free, no account required)
• Connect external server — point Manufact’s web UI at any running MCP server URL to get the Chat interface, test suites, and publishing checks without deploying there
• Deploy to Manufact — adds GitHub-triggered deploys, preview environments, and gateway analytics

FAQ

Q: Is mcp-use open source? A: Yes. The TypeScript and Python SDKs (servers, clients, and agents) are MIT-licensed and open source. The Manufact Cloud platform builds on them.

Q: Can I self-host an MCP server built with mcp-use? A: Yes. mcp-use generates standard MCP servers that you can host anywhere. The “deploy to Manufact” step is optional.

Q: What clients support MCP? A: ChatGPT (via ChatGPT Apps), Claude Desktop (via the MCP protocol), and any other client implementing the Model Context Protocol specification.

Q: Does Manufact support Python MCP servers? A: Yes. mcp-use has SDKs for both TypeScript and Python.

Q: What is the “vibecode” feature mentioned on the site? A: The “Vibecode” flow lets you describe the MCP server you want in natural language and watch it scaffold in real time on the Manufact Cloud platform.

Conclusion

If you are building anything that exposes tools or data to AI agents — whether a customer-facing ChatGPT App, a Claude connector for your internal APIs, or a general-purpose MCP server — mcp-use gives you a typed, well-structured foundation. The Manufact Cloud layer then handles the operational overhead of testing across clients, auditing for store compliance, and monitoring in production. The open-source SDK means you are never locked in; the cloud platform means you can ship faster without reinventing deployment and testing infrastructure.

For a full walkthrough, see the mcp-use documentation or try the cloud Inspector against a live MCP server.