What problem does it solve? Lambda-style serverless functions time out on long-running workloads like AI agent streams, WebSocket servers, and SSE endpoints. This Skill guides you through deploying long-running Node.js 24 HTTP handlers directly onto a Neon branch, with DATABASE_URL injected automatically and compute running in the same region as your Postgres data. ## Core Features & Use Cases - Long-Running Handlers: Deploy web-standard fetch handlers with a 15-minute time-to-first-byte budget, supporting streaming agents, WebSocket servers, and SSE endpoints that outlast traditional serverless limits. - Branch-Scoped Backends: Each Neon branch runs its own function version at its own URL against its own isolated database, so preview and CI environments get self-contained backends via neon.ts infrastructure-as-code. - Integrated References: Includes in-depth guides for building Vercel AI SDK agents, Mastra agents with observability, MCP servers, SSE endpoints, and Sentry error monitoring on Functions. - Use Case: You have a Next.js app on Vercel whose AI agent stream gets cut off by the host's 60-second limit. Move just the agent onto a Neon Function, mint a short-lived JWT on your backend, and have the browser call the function directly so the stream runs to completion next to your database. ## Quick Start Ask the AI to define a Hono-based function in neon.ts, run neon dev locally, and deploy it with neon deploy to get a public invocation URL.
