Most of the conversation about AI agents focuses on what they can do. Very little of it focuses on the plumbing that makes them work.

Manufact, a Y Combinator Summer 2025 company formerly known as mcp-use, is betting that the plumbing is where the real value is. In February 2026, the company raised $6.3 million in seed funding led by Peak XV, with participation from Liquid 2 Ventures, Ritual Capital, Pioneer Fund, and Y Combinator itself.

The product Manufact builds is an open-source SDK and cloud hosting layer for Model Context Protocol servers: the infrastructure that lets AI agents connect reliably to external tools and data sources at production scale.

The Problem Manufact Solves

When an AI agent needs to access a database, read a file, call a Slack API, or interact with any external system, it needs a bridge. The Model Context Protocol defines how that bridge is supposed to work. But defining a protocol and running a protocol in production are two different things.

Most developers who want to deploy MCP servers are not infrastructure engineers. They do not want to manage server provisioning, handle uptime, configure authentication flows, or deal with the operational complexity of running networked services at scale. They want to define what their AI agent should be able to do and have infrastructure handle the rest.

That is the gap Manufact fills. The open-source SDK handles the protocol implementation. The cloud hosting layer handles the operational complexity: deployment, scaling, monitoring, and reliability.

The comparison that helps frame the value proposition is what Vercel did for Next.js or what Heroku did for Ruby on Rails. Both frameworks were usable before those hosting platforms existed. Both became dramatically more accessible after. Manufact is making a similar bet that MCP will follow the same trajectory.

Why Peak XV and YC Backed This

Peak XV, formerly Sequoia India and Southeast Asia, is not a fund that bets small. When it leads a $6.3 million seed round in an infrastructure company, it is expressing a view about the addressable market.

The market thesis here is that MCP becomes a durable standard across the AI industry, and that the hosting and tooling layer for MCP servers becomes as commercially significant as the hosting layers for any major internet protocol. If every enterprise deploying AI agents eventually runs MCP servers as part of their architecture, and if Manufact is the easiest way to do that, the market is very large.

Y Combinator’s participation is a signal on execution. YC does not co-invest in seed rounds it backed through the batch without confidence in the team.

The name change from mcp-use to Manufact is also a deliberate positioning choice. mcp-use was a descriptive name for an open-source tool. Manufact is a brand for a commercial infrastructure company. That shift reflects an understanding that the opportunity is not just in building developer tools but in owning a layer of the enterprise AI stack.

What This Means for Businesses Deploying AI Agents

If your organisation is deploying AI agents that need to connect to internal systems, the decision about MCP infrastructure is one you will need to make.

The options are building your own MCP server infrastructure, running open-source tooling on your own compute, or using a hosted service like what Manufact is building. For most businesses, the build-your-own path is unnecessarily expensive unless you have specific security or compliance requirements that prevent you from using external hosting.

The more interesting question is how quickly MCP becomes a standard expectation in enterprise AI procurement. If the tools and platforms your business buys start shipping MCP support as a default, the question shifts from “should we adopt MCP?” to “what is our strategy for managing MCP connections securely across our agent deployments?”

Manufact’s $6.3 million round is early. But the investors behind it are not making early bets. They are backing a thesis about where the infrastructure layer of enterprise AI is going to sit, and doing it while the market is still forming.

---

Enterprise DNA helps businesses design and deploy AI agent systems that connect to the data and tools you already rely on. Talk to us about your AI architecture.