Neon Functions

Neon Functions

Neon Functions are long-running Node.js HTTP handlers deployed onto a Neon branch. Each function gets a public HTTPS URL, runs in the same region as your database (us-east-2), and — if the branch has Postgres — gets DATABASE_URL injected automatically. Deploy and manage them through the same Neon CLI, neon.ts, and API you already use.

Preview feature. Only available on new projects in us-east-2. Cannot be enabled on existing projects. Usage is not billed during the private preview. If the user lacks access, point them to: https://neon.com/blog/were-building-backends#access

When to Use

Reach for Neon Functions when the workload is a request/response handler that benefits from staying alive and staying close to the data:

• Long-running request/response flows that outlast lambda-style limits (~10–60s). Agents making multiple LLM calls, image/video generation, etc. The handler must start responding within 15 minutes; an open stream stays alive as long as bytes keep flowing.
• Stateful streaming — SSE endpoints or WebSocket servers held open in-process. No external state store (Redis) needed. Module-scope state (a pg pool, in-memory counter) persists across requests on the same isolate.
• Compute next to Postgres — runs in the branch's region; no cross-region round trips. DATABASE_URL injected for you.
• A backend that branches with your data — each branch runs its own function version at its own URL against its own isolated database. Preview deployments, CI, and dev environments each get a self-contained backend.
• Webhooks, bots, post-response work — fan-out DB writes, Discord/WebSocket bots, fire-and-forget follow-ups via waitUntil.

Do NOT use when: the workload is a pure static site, a cron/background job needing its own lifecycle and cancellation, or anything that must run outside us-east-2 today.

Prerequisites

• A new Neon project in us-east-2 (Functions cannot be enabled on existing projects).
• Neon CLI installed and authenticated (neon command available).
• Node.js 24 runtime (Functions run on Node.js 24; memory fixed at 2048 MiB during preview).
• @neon/config package for neon.ts infrastructure-as-code.
• For Postgres-connected functions: the branch must have Postgres enabled (otherwise DATABASE_URL is not injected).

Procedure

1. Declare functions in neon.ts

Add @neon/config and declare functions under preview.functions, keyed by slug:

// neon.ts
import { defineConfig } from "@neon/config/v1";

export default defineConfig({
  preview: {
    functions: {
      todos: {
        // slug: ^[a-z0-9]{1,20}$ — lowercase letters/digits, no hyphens
        name: "todo api", // display label only
        source: "src/index.ts", // entry file, relative to neon.ts
      },
    },
  },
});

HARD RULE: The slug is the function's permanent identity (appears in the invocation URL and CLI commands) and cannot be changed after the first deploy. Use name for a human-readable label. Slugs must match ^[a-z0-9]{1,20}$.

2. Write the function handler

A function is any default export with a fetch(request) method returning a Response (Workers/WinterTC-compatible). A Hono app exports exactly that shape, so export default app just works.

Minimal example — Hono app querying Postgres via injected DATABASE_URL:

// src/index.ts
import { Hono } from "hono";
import { drizzle } from "drizzle-orm/node-postgres";
import { Pool } from "pg";
import { parseEnv } from "@neon/env";
import config from "../neon";
import { todos } from "./db/schema";

const env = parseEnv(config);
const pool = new Pool({ connectionString: env.postgres.databaseUrl, max: 5 });
const db = drizzle(pool);

const app = new Hono();
app.get("/", (c) => c.text("Neon + Hono + Drizzle"));
app.post("/todos", async (c) => {
  const { text } = await c.req.json<{ text: string }>();
  const [row] = await db.insert(todos).values({ text }).returning();
  return c.json(row, 201);
});
app.get("/todos", async (c) => c.json(await db.select().from(todos)));

export default app;

HARD RULE: Create the pg pool once at module scope and reuse it across requests. Do NOT open a connection per request. Keep max small (e.g. 5) — each isolate keeps its own pool, so total connections scale with the number of live isolates.

HARD RULE: Use Drizzle (or another ORM) on top of node-postgres (pg) — not Neon's serverless driver. Functions are long-running and reuse an isolate across many requests, so a persistent pg pool is the right fit.

To scope parseEnv to only the keys you need:

const { postgres } = parseEnv(config, ["DATABASE_URL"]);
const pool = new Pool({ connectionString: postgres.databaseUrl, max: 5 });

Reading process.env.DATABASE_URL directly also works everywhere.

3. Develop locally

neon dev      # serves every function in neon.ts with hot reload; injects DATABASE_URL & friends

neon dev injects NEON_BRANCH, DATABASE_URL, DATABASE_URL_UNPOOLED, and other Neon-managed vars into your local dev environment so local runs mirror the deployed runtime.

4. Deploy

neon deploy   # bundles with esbuild, uploads, and applies neon.ts to the linked branch

Deploy a single function without neon.ts:

neon functions deploy <slug> --path . --entry src/index.ts

Retrieve the public URL:

neon functions get <slug>
# Look for the invocation_url field:
# https://<branch_id>-<slug>.compute.c-1.us-east-2.aws.neon.tech

Manage functions:

neon functions list
neon functions get <slug>
neon functions delete <slug>

5. Infrastructure-as-code with neon.ts

neon config status   # print the branch's live config (deployed functions)
neon config plan     # dry-run diff of what apply would change
neon config apply    # bundle + deploy the declared functions (neon deploy is an alias)

HARD RULE: Functions are branch-scoped. When neon.ts is present, neon checkout applies the policy as it creates a branch. Checking out an existing branch does not re-deploy — run neon deploy explicitly.

Per-branch deploy tuning (e.g. runtime) lives in the branch closure:

export default defineConfig({
  preview: {
    functions: { todos: { name: "todo api", source: "src/index.ts" } },
  },
  branch: (branch) => ({
    preview: { functions: { todos: { runtime: "nodejs24" } } },
  }),
});

6. Environment variables

Neon-injected (automatic, do not declare or pass at deploy time):

Object storage (AWS_*) and AI Gateway (OPENAI_*, NEON_AI_GATEWAY_*) vars are also injected when those services are declared.

Your own secrets — set with --env KEY=VALUE on deploy (repeatable; --env KEY= deletes a key; unmentioned keys carry over), or declare in neon.ts:

functions: {
  todos: {
    name: "todo api",
    source: "src/index.ts",
    env: { OPENAI_API_KEY: process.env.OPENAI_API_KEY! },
  },
}

Load a .env before deploy: neon deploy --env .env.production

Pull branch vars for local dev: neon env pull (link/checkout do this automatically; pass --no-env-pull to skip).

HARD RULE — Env limits: ≤1,000 vars, ≤64 KiB total, and the NEON_ prefix is reserved.

7. Connecting to Postgres

• DATABASE_URL — pooled (through Neon's connection pooler). Use for normal request/response traffic. Every Postgres ORM reads DATABASE_URL by default.
• DATABASE_URL_UNPOOLED — direct. Use for migrations, LISTEN/NOTIFY, long multi-statement transactions.

import { drizzle } from "drizzle-orm/node-postgres";
import { Pool } from "pg";

const pool = new Pool({ connectionString: process.env.DATABASE_URL, max: 5 });
const db = drizzle(pool);

You do NOT need to close the pool on shutdown — when the runtime evicts an isolate it sends SIGINT/SIGTERM, and Neon's pooler reclaims connections automatically.

8. WebSocket servers

A function's default export is normally { fetch }. To accept WebSockets, export an upgrade method alongside it — the runtime routes plain HTTP to fetch and the WebSocket handshake to upgrade:

export default {
  fetch(request: Request): Response | Promise<Response> { /* HTTP */ },
  async upgrade(req: IncomingMessage, socket: Duplex, head: Buffer) { /* WS handshake */ },
};

Simple example with ws library and token auth:

// src/index.ts
import type { IncomingMessage } from "node:http";
import type { Duplex } from "node:stream";
import { WebSocketServer, type WebSocket } from "ws";

const clients = new Set<WebSocket>();
const wss = new WebSocketServer({ noServer: true });

export default {
  fetch: () => new Response("WebSocket endpoint — connect with ?token=<jwt>"),
  async upgrade(req: IncomingMessage, socket: Duplex, head: Buffer) {
    const url = new URL(req.url ?? "/", "http://localhost");
    const identity = await verifyToken(url.searchParams.get("token"));
    if (!identity) {
      socket.write("HTTP/1.1 401 Unauthorized\r\n\r\n");
      socket.destroy();
      return;
    }
    wss.handleUpgrade(req, socket, head, (ws) => {
      clients.add(ws);
      ws.on("close", () => clients.delete(ws));
      ws.on("message", (data) => broadcast(data.toString()));
    });
  },
};

HARD RULE — Heartbeat: A connection stays open only while bytes flow. Neon evicts a silent stream after 15 minutes, and intermediary proxies are often stricter (tens of seconds). Send a periodic ping:

const HEARTBEAT_MS = 25_000;

const beat = setInterval(() => {
  for (const ws of clients) if (ws.readyState === ws.OPEN) ws.ping();
}, HEARTBEAT_MS);
beat.unref?.();

HARD RULE — Fan-out across isolates: Under load, the runtime runs several isolates in parallel, each with its own copy of module state. An in-memory Set<WebSocket> only covers the current isolate. For true fan-out (broadcasting to all clients across all isolates), use Postgres LISTEN/NOTIFY or a shared pub/sub layer. Do NOT skip this if you need broadcast semantics.

HARD RULE — Client must reconnect: The platform may evict/restart for operational reasons (maintenance, moving compute). Treat eviction like a process restart — WebSocket/SSE clients must reconnect. The platform sends SIGINT before evicting.

HARD RULE — No header-modifying middleware on upgradeWebSocket routes: CORS or similar middleware that rewrites headers will throw when combined with Hono's upgradeWebSocket() helper.

When to load reference files:

• references/hono-websockets.md — Load when the user wants to declare WebSocket routes inside a Hono app using app.get("/ws", upgradeWebSocket(...)) with onOpen/onMessage/onClose lifecycle. Contains a self-contained createNeonWebSocket(app) adapter (depends only on hono and ws; no deprecated @hono/node-ws package).

9. Server-sent events (SSE)

SSE works naturally — return a Response with Content-Type: text/event-stream and keep writing chunks. The stream stays alive as long as bytes flow (15-minute heartbeat applies).

10. Functions as an agent backend

HARD RULE — Never proxy a long agent stream through your app backend. Platforms like Vercel/Netlify/Cloudflare cap serverless/edge execution at ~10–60s (sometimes up to ~300s). A long agent stream gets cut off mid-response even though the Neon Function would keep going.

Browser ──(Authorization: Bearer <JWT>)──▶  Neon Function (agent)   ✅ no host timeout
Browser ──▶ your app backend ──▶ Neon Function                       ❌ host cuts the stream

HARD RULE — A Neon Function has a public HTTPS URL reachable by anyone. A direct client→function call means no app backend gates access. You must authenticate the function yourself: verify a JWT (against your app's JWKS), check a shared secret/API key, or validate a session token at the top of the handler. Never deploy an unauthenticated agent.

// src/index.ts — verify the caller before doing any work
import { createRemoteJWKSet, jwtVerify } from "jose";

const jwks = createRemoteJWKSet(new URL(`${process.env.AUTH_BASE_URL}/api/auth/jwks`));

export default {
  async fetch(request: Request) {
    if (request.method === "OPTIONS") return new Response(null, { status: 204, headers: cors(request) });

    const auth = request.headers.get("authorization");
    if (!auth?.toLowerCase().startsWith("bearer ")) {
      return new Response("Unauthorized", { status: 401, headers: cors(request) });
    }
    try {
      const { payload } = await jwtVerify(auth.slice(7), jwks, {
        issuer: process.env.AUTH_BASE_URL,
        audience: process.env.AUTH_BASE_URL,
      });
      const userId = payload.sub; // scope the agent to this user
      // ... run the agent, return result.toUIMessageStreamResponse({ headers: cors(request) })
    } catch {
      return new Response("Unauthorized", { status: 401, headers: cors(request) });
    }
  },
};

When to load reference files:

• references/ai-sdk.md — Load when the user wants a complete Vercel AI SDK agent running as a Function (streaming toUIMessageStreamResponse, multi-step tool calling next to Postgres, persisting generated images to Object Storage).
• references/mastra-studio.md — Load when the user wants to run a Mastra agent on a function and ship traces to Mastra Studio (Mastra Cloud) for observability.
• references/sentry.md — Load when the user needs error monitoring/observability for a long-running function.

Pitfalls

1. Slug is permanent. The slug appears in the invocation URL and CLI commands and cannot be changed after the first deploy. Choose carefully. Must match ^[a-z0-9]{1,20}$ — lowercase letters/digits only, no hyphens.

2. Opening a DB connection per request. This exhausts Postgres connections under load. Create the pg pool once at module scope with max: 5 and reuse it.

3. Using Neon's serverless driver instead of pg. Functions are long-running and reuse isolates — a persistent pg pool is the correct choice, not the HTTP-based serverless driver.

4. No heartbeat on WebSocket/SSE. Silent streams are evicted after 15 minutes (proxies often sooner, tens of seconds). Send a ping every ~25 seconds.

5. Assuming module state is shared across isolates. Under load, multiple isolates run in parallel, each with its own copy of module state. In-memory data structures (like a Set<WebSocket>) are per-isolate. Persist anything that must survive eviction or be shared in Postgres.

6. Proxying agent streams through the app backend. Host platforms cut the stream at their serverless limits. Call the Neon Function directly from the client with a short-lived JWT.

7. Deploying an unauthenticated function. The public HTTPS URL is reachable by anyone. Always verify a JWT/API key/session token at the top of the handler.

8. Forgetting CORS for direct client→function calls. Handle OPTIONS preflight and set Access-Control-Allow-Origin/-Headers so the browser can reach the function cross-origin.

9. Expecting neon checkout to redeploy on existing branches. It only applies the policy when creating a branch. Run neon deploy explicitly after checking out an existing branch.

10. Using NEON_ prefix for your own env vars. This prefix is reserved. Also respect limits: ≤1,000 vars, ≤64 KiB total.

11. Putting CORS/header-modifying middleware on upgradeWebSocket routes. The helper rewrites headers internally and will throw.

12. Not treating eviction as a process restart. The platform may evict/restart for operational reasons. WebSocket/SSE clients must reconnect. Use process.on("SIGINT", ...) for last-minute cleanup if needed (not needed for closing Postgres connections — the pooler reclaims those).

Verification

1. Check function is deployed and get its URL:

   neon functions get <slug>
   

   Expected output includes invocation_url of the form:

   https://<branch_id>-<slug>.compute.c-1.us-east-2.aws.neon.tech
   

2. List all deployed functions:

   neon functions list
   

3. Check live config matches neon.ts:

   neon config status
   

4. Dry-run a deploy to see what would change:

   neon config plan
   

5. Test the function endpoint:

   curl -s https://<branch_id>-<slug>.compute.c-1.us-east-2.aws.neon.tech/
   

   Expected: Neon + Hono + Drizzle (or your handler's response).

6. Verify local dev serves the function:

   neon dev
   # Then curl the local URL printed in the output
   

7. Verify against official docs (source of truth, Functions is evolving rapidly):

   • Docs index: https://neon.com/docs/llms.txt
   • Overview: https://neon.com/docs/compute/functions/overview.md
   • Get started: https://neon.com/docs/compute/functions/get-started.md
   • Deploy: https://neon.com/docs/compute/functions/deploy.md
   • Environment variables: https://neon.com/docs/compute/functions/environment-variables.md
   • neon.ts reference: https://neon.com/docs/compute/functions/reference/neon-ts.md
   • Runtime limits: https://neon.com/docs/compute/functions/reference/runtime-limits.md
   • Preview access: https://neon.com/docs/compute/functions/preview-access.md

   Any doc page can be fetched as markdown by appending .md to the URL or requesting Accept: text/markdown.

Timeouts and Runtime Limits

An isolate is reused across many requests — multiple requests can be in flight on the same isolate at once (interleaved on Node's single-threaded event loop). Under load, several isolates run in parallel, each with its own copy of module state. State held in module scope is per-isolate and in-memory only.

Related skills

• neon — Branch-first workflow and neon.ts basics.
• neon-object-storage — Object storage service (AWS_* env vars injected when declared).
• neon-ai-gateway — AI Gateway service (OPENAI_*, NEON_AI_GATEWAY_* env vars; one credential across every model).