MCP server support on Vercel
Vercel just made it dead simple to deploy AI agent toolkits. One customer cut costs by 90%.

Why it matters
Vercel's MCP server support lowers the friction for building and deploying AI agent integrations at scale. This matters because agentic workflows are moving from proof-of-concept to production, and infrastructure that handles stateless HTTP transport + cost-efficient compute is becoming table stakes.
The key facts
6 to knowVercel now supports deploying MCP servers and clients natively
New @vercel/mcp-adapter package supports both SSE and stateless HTTP transport
One customer achieved 90% cost savings using Vercel Fluid compute vs traditional serverless for AI inference
MCP protocol now supports HTTP + OAuth (improvement from stateful SSE)
Vercel AI SDK has built-in MCP support for Node.js/Next.js apps
Redis integration available via marketplace (Upstash, Redis Labs) for SSE state management
Go to the source
Vercel Blogvercel.com
Publisher excerpt: is a way to build integrations for AI models.Model Context Protocol (MCP) Vercel now supports deploying MCP servers (which AI models can connect to) as well as MCP clients ( applications which call the servers).AI chatbot Get started with our today.Next.js MCP template APIs allow different services…