Introducing mcp-c: The Cloud Platform for AI Agents

Building powerful AI agents is one thing; deploying them is another. Unlike traditional web applications, agents often need to manage long-running tasks, maintain state, and handle complex operations that can pause and resume. LastMile AI has just launched mcp-c, a new cloud platform in open beta built specifically to solve these AI-native deployment problems.

This new service is designed for hosting mcp-agent applications, MCP servers, and even ChatGPT apps. The goal is to provide a seamless "local to cloud" experience for developers, similar to what Vercel offers for Next.js applications.

Why a New Platform?

mcp-c is built around the Model Context Protocol (MCP), a framework for creating effective AI agents. The platform addresses the unique runtime issues these agents face, such as the need for fault-tolerant, long-running operations. Standard cloud platforms often fall short here, requiring significant custom infrastructure.

The platform provides this robust foundation out-of-the-box, allowing developers to focus on agent behavior rather than complex deployment wiring. It's free to use during the open beta, offering a chance to test its capabilities.

Durable Execution and Easy Deployment

The standout feature is durable execution via Temporal. This allows your agents to run complex, long-running tasks that can pause, resume, and survive failures without losing state. This is essential for agents that might need to run for minutes or even hours to complete a user's request.

Getting started is designed to be fast. The team at LastMile AI emphasizes a quick setup, providing a 5-minute quickstart. You can initialize, test, and deploy an agent with just a few CLI commands.

uvx mcp-agent init
uv init
uv add "mcp-agent[openai]"
# Set your provider key (OpenAI / Anthropic etc.)
uv run main.py # test locally (optional)
uvx mcp-agent login
uvx mcp-agent deploy --no-auth
 Copy

Ready for Production

Beyond simple deployment, mcp-c is built to be production-ready. It supports the entire MCP specification, which includes critical features like authentication with a managed OAuth server, observability hooks, and agent-specific patterns like sampling and elicitation.

This platform aims to bridge the gap between building a clever AI prototype and running a reliable, secure AI service. By handling the complex infrastructure, mcp-c lets developers focus on what their agents can do, not just how they stay online.