Manufact has raised a $6.3M seed round to help developer teams build and deploy MCP servers for AI agents and MCP apps for ChatGPT and Claude.
The round is led by Peak XV — formerly Sequoia Capital India and Southeast Asia, now managing over $10B in assets — with participation from Y Combinator, Liquid 2 Ventures, Ritual Capital, Pioneer Fund, True Ventures, Transpose Platform, Italian Founders Fund, ACE Ventures, Vento, and Spot VC, plus a deep bench of operator angels: Kulveer Taggar (Phosphor Capital), Ant Wilson (co-founder & COO, Supabase), Rory Wilding, Eli Brown, Benjamin Bryant, Henri Stern, Jake Klamka, Mike Hinckley, Pancrazio Auteri, Nicole Zingg, Michael Mason, Alec Howard, and others.
The bet: software is going to be built for agents
For decades, every button, menu, and dashboard existed to translate a human's intention into a machine's action. That era is ending. As AI agents take over the work humans do inside software — filing expense reports, managing tickets, writing code, booking travel — every product on earth will need a new kind of interface designed for the agent, not the user.
"Software products are already being accessed by, and will be accessed mainly by, AI agents — or by users through chat interfaces. That's our bet. That's our thesis. And that's what we are really rooting our company on." — Luigi Pederzani, co-founder & co-CEO
The shift is the same one mobile went through fifteen years ago. Most products are still being built web-first, with an MCP server bolted on as an afterthought. Manufact's view is that the next generation of software will be MCP-first and chat-first — and the companies that don't make their products accessible to agents risk being reduced to systems of record that agents query but no longer own the user experience for.
Why MCP — the "USB-C for AI"
Manufact is built on top of the Model Context Protocol (MCP), the open standard introduced by Anthropic in late 2024 that has rapidly become the dominant way for AI agents to talk to external tools and data.
Before MCP, connecting an agent to a company's software meant bespoke connectors for every tool — fragile, expensive, slow. MCP collapsed that into one standard: a universal connector, the USB-C for AI.
The adoption curve has been brutal in the best way:
- In December 2025, Anthropic donated MCP to the Linux Foundation's new Agentic AI Foundation, co-founded with Block and OpenAI, with support from Google, Microsoft, AWS, and Cloudflare.
- 10,000+ active public MCP servers are in production today.
- ~7M MCP server downloads per month across the ecosystem.
- ChatGPT, Cursor, Google Gemini, Microsoft Copilot, and Visual Studio Code all support the protocol.
- The global AI agents market reached $7.84B in 2025 and is projected to hit $52.62B by 2030.
From mcp-use to Manufact
Manufact was founded as mcp-use by Pietro Zullo and Luigi Pederzani — two Italians who met at a Zurich co-working space (the same one that produced Browser Use, Bloom, and several other YC companies). Pietro was studying at ETH Zurich; Luigi was at Morgen, an ETH spin-off used by teams at Spotify, GitHub, and Linear, after leading a 12-engineer team at Accenture Switzerland.
Both were winding down previous projects when MCP launched. They had each written agents before and remembered how painful the tool/integration plumbing was.
"When MCP came out, it looked like the perfect fit for what we were trying to do. But only Cursor, Claude Code, and a few closed-source apps actually let you use the protocol. I'm not going to do my groceries from Cursor. So we wrote an open-source library to do what Cursor was doing — but on your own machine, in your own app, on your own terms." — Pietro Zullo, co-founder
They called it mcp-use, with a tagline that did most of the marketing for them: "Connect any MCP to any LLM in six lines of code." The repo picked up 2,000+ GitHub stars within weeks. They applied to Y Combinator's Summer 2025 batch on the day of the deadline.
Today, the open-source SDKs have crossed 6M+ downloads and 9,000+ GitHub stars. Organizations including NASA, Nvidia, and SAP use the libraries, and roughly 20% of the US 500 have experimented with them.
What Manufact ships
The strategy borrows from the playbook that turned Vercel into a multi-billion-dollar company — but for the agent-native layer. Manufact ships three tightly integrated products:
- mcp-use SDK (open source, Python + TypeScript) — spin up a fully functional AI agent connected to MCP tools in as few as six lines of code. Works with any LLM, including local models, and integrates with LangChain and friends.
- mcp-use Inspector — a built-in browser-based debugger and testing suite. Inspect raw JSON-RPC traffic, hot-reload servers, and test tool execution in a sandbox without wiring up a live agent.
- Manufact Cloud — deploy, scale, auth, access control, observability. Connect a GitHub repo and you go from
git pushto a production MCP server in under 60 seconds, instantly reachable from ChatGPT, Claude, and the broader agent ecosystem.
Manufact has also moved aggressively into MCP apps — the newer extension of the protocol that lets developers render interactive React widgets, data visualizations, and forms directly inside chat clients. The SDK scaffolds an MCP app from a single terminal command, and the same one-click deploy puts it in front of the 800M+ users of ChatGPT and Claude.
"As software becomes more agentic, the hard part isn't the model anymore — it's everything around it. We started Manufact because developers were spending too much time on plumbing instead of building and shipping their products." — Pietro Zullo
A community-first company
Mcp-use grew out of an open-source community before it was a company, and Manufact is determined to keep it that way. After YC, the team made the deliberate call to focus entirely on the open-source product and community rather than rushing monetization.
In late February, Manufact co-hosted what was, at the time, the largest MCP apps hackathon ever — at YC's headquarters in San Francisco. 650 applications, 300 builders, sponsored by OpenAI, Cloudflare, and Anthropic. Eight Anthropic employees showed up — more people than Manufact's entire team.
What's next
The capital goes straight into the roadmap:
- Harden the SDK and Inspector so every MCP server feels production-grade by default — auth, observability, versioning, evals.
- Deepen Manufact Cloud — faster deploys, richer observability, first-class auth, and one-click distribution to the major AI clients.
- Grow the open-source surface so that whatever the protocol does next, the easiest place to build on it stays the same.
- Ship MCP apps as a first-class distribution channel — React widgets inside ChatGPT and Claude, reachable by hundreds of millions of users from day one.
The long-term metric the team is optimizing for is straightforward: the share of global AI tool calls that flow through Manufact.
"Like Stripe is doing for global GDP — we're going to win if we can get a great number for it." — Luigi Pederzani
Thanks to every developer who has shipped an MCP server on Manufact, every contributor to the open-source SDKs, and every investor who put conviction behind this round. The bet is big and the timing is right.
If you're building agents, building MCP servers, or building products that should be reachable from inside ChatGPT and Claude — come build with us.
