From netlify-skills
Guides writing Netlify serverless functions with modern syntax, TypeScript, path routing, background/scheduled tasks, streaming, and method routing.
How this skill is triggered — by the user, by Claude, or both
Slash command
/netlify-skills:netlify-functionsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Always use the modern default export + Config pattern. Never use the legacy `exports.handler` or named `handler` export.
Always use the modern default export + Config pattern. Never use the legacy exports.handler or named handler export.
import type { Context, Config } from "@netlify/functions";
export default async (req: Request, context: Context) => {
return new Response("Hello, world!");
};
export const config: Config = {
path: "/api/hello",
};
The handler receives a standard Web API Request and returns a Response. The second argument is a Netlify Context object.
The bare default export shown above is the recommended form. The default export can also be an object with a fetch method — prefer this form only if (1) other functions in the project already use it, or (2) the function also subscribes to platform events (see Event Handlers below), since events are exposed as named handlers on the same object:
export default {
fetch(req: Request, context: Context) {
return new Response("Hello, world!");
},
};
Place functions in netlify/functions/:
netlify/functions/
_shared/ # Non-function shared code (underscore prefix)
auth.ts
db.ts
items.ts # -> /.netlify/functions/items (or custom path via config)
users/index.ts # -> /.netlify/functions/users
Use .ts or .mts extensions. If both .ts and .js exist with the same name, the .js file takes precedence.
Define custom paths via the config export:
export const config: Config = {
path: "/api/items", // Static path
// path: "/api/items/:id", // Path parameter
// path: ["/api/items", "/api/items/:id"], // Multiple paths
// excludedPath: "/api/items/special", // Excluded paths
// preferStatic: true, // Don't override static files
};
Without a path config, functions are available at /.netlify/functions/{name}. Setting a path makes the function available only at that path.
Access path parameters via context.params:
// config: { path: "/api/items/:id" }
export default async (req: Request, context: Context) => {
const { id } = context.params;
// ...
};
export default async (req: Request, context: Context) => {
switch (req.method) {
case "GET": return handleGet(context.params.id);
case "POST": return handlePost(await req.json());
case "DELETE": return handleDelete(context.params.id);
default: return new Response("Method not allowed", { status: 405 });
}
};
export const config: Config = {
path: "/api/items/:id",
method: ["GET", "POST", "DELETE"],
};
For long-running tasks (up to 15 minutes). The client receives an immediate 202 response; return values are ignored.
Enable background mode by setting background: true in config:
export default async (req: Request) => {
await someLongRunningTask();
};
export const config: Config = {
path: "/process",
background: true,
};
Store results externally (Netlify Blobs, database) for later retrieval.
The legacy filename convention (process-background.ts) is still supported, but new functions should use config.background.
Functions deploy to cmh (Ohio) by default for new sites. This is a deliberate choice: US East is centrally located for an international audience, has a broad provider ecosystem, and gives most projects the lowest overall latency without any configuration. Sites created before October 4, 2023 may have a different default region — check Project configuration > Build & deploy > Continuous deployment > Functions region to confirm.
Do NOT override config.region unless the user has stated a specific reason — for example, a database or backend service in another region with measurable roundtrip savings, a data-residency requirement, or an audience concentrated in one region whose compute dependencies (database, backend services) also live in that region.
Two constraints to be aware of before adding config.region:
config.region in code. The generated files can't carry per-function config.Region values are IATA airport codes — for example cmh (Columbus/Ohio, the default), dub (Dublin), and fra (Frankfurt). See Region for the full list of supported regions and details.
Functions run with 1024 MB of memory and a proportional amount of compute by default. The default fits most workloads, and raising it has a direct cost impact: function billing scales linearly with the configured size.
Do NOT set config.memory or config.vcpu speculatively. Only reach for them when:
memory and vcpu configure the same underlying resource and are mutually exclusive — set one, not both. Set memory as a number of megabytes (e.g. memory: 2048) or as a string with a unit (e.g. '2gb' or '2048mb', case-insensitive), within the 1024–4096 MB range. Adjusting them is available only on Credit-based Pro and Enterprise plans; on other plans these settings have no effect. See Memory or vCPU for accepted values and the exact mapping.
Run on a cron schedule (UTC timezone):
export default async (req: Request) => {
const { next_run } = await req.json();
console.log("Next invocation at:", next_run);
};
export const config: Config = {
schedule: "@hourly", // or cron: "0 * * * *"
};
Shortcuts: @yearly, @monthly, @weekly, @daily, @hourly. Scheduled functions have a 30-second timeout and only run on published deploys.
Testing and triggering. A scheduled function does not fire on its cron schedule under netlify dev — the local dev server never runs the schedule, so waiting for the clock will appear to do nothing. Test it by invoking it directly: netlify functions:invoke <name> calls the function once, on demand. In production a scheduled function also has no public HTTP URL — it is not reachable at /.netlify/functions/{name} and cannot be triggered by an external HTTP request; it runs only on its schedule. If you also need to trigger the same work over HTTP (a manual "run now" or a webhook), expose that logic through a separate ordinary HTTP function and share the implementation rather than trying to POST to the scheduled function.
Return a ReadableStream body for streamed responses (up to 20 MB):
export default async (req: Request) => {
const stream = new ReadableStream({ /* ... */ });
return new Response(stream, {
headers: { "Content-Type": "text/event-stream" },
});
};
A function can subscribe to platform events by exporting an object instead of a function as its default. Each event has a named handler property:
import type { DeploySucceededEvent, UserSignupEvent } from "@netlify/functions";
export default {
deploySucceeded(event: DeploySucceededEvent) {
console.log(`Deploy ${event.deploy.id} succeeded`);
},
userSignup(event: UserSignupEvent) {
return {
user: {
...event.user,
appMetadata: { ...event.user.appMetadata, roles: ["member"] },
},
};
},
};
A single function can declare multiple handlers; multiple functions can also subscribe to the same event.
Available handlers:
| Handler | Trigger |
|---|---|
fetch | HTTP request (equivalent to a bare function default export) |
deployBuilding / deploySucceeded / deployFailed / deployDeleted / deployLocked / deployUnlocked | Deploy lifecycle |
userSignup / userLogin / userValidate / userModified / userDeleted | Identity lifecycle |
formSubmitted | Form submission verified |
userSignup, userLogin, userValidate, and userModified can reject the action by calling event.deny(). The end user receives a 401; no observability error is produced (unlike throwing).
export default {
userLogin(event: UserLoginEvent) {
if (!event.user.email?.endsWith("@example.com")) {
return event.deny();
}
},
};
If multiple functions subscribe to the same event, the first to call event.deny() aborts the chain.
| Property | Description |
|---|---|
context.params | Path parameters from config |
context.geo | { city, country: {code, name}, latitude, longitude, subdivision, timezone, postalCode } |
context.ip | Client IP address |
context.cookies | .get(), .set(), .delete() |
context.deploy | { context, id, published } |
context.site | { id, name, url } |
context.account.id | Team account ID |
context.requestId | Unique request ID |
context.waitUntil(promise) | Extend execution after response is sent |
context.geo and context.ip are mocked under netlify dev. Locally these return placeholder values, not your real location or client IP, so a value that looks "stuck" on a default country does not mean your geo code is broken — real geolocation is populated only for deployed functions. To exercise geo branching locally, start dev with the geo flags: netlify dev --geo=mock --country=DE forces mock data (--geo=mock) and sets the mock country (--country). Don't conclude context.geo is broken because local values never change.
Prefer Netlify.env.get inside functions:
const apiKey = Netlify.env.get("API_KEY");
process.env is also valid inside Functions and reads the same variables — prefer Netlify.env.get for cross-runtime and edge portability (a function you later move to an Edge Function keeps working, since Edge Functions expose only Netlify.env.get, not process.env).
Environment variables have a small total size budget. Functions run on AWS Lambda, which caps the combined size of all environment variables at roughly 4 KB. A single large value — a service-account JSON credential, a PEM private key, a big config blob — can blow past that on its own and break the deploy or the function at runtime. Do not store large payloads in environment variables; keep only small secrets and config (API keys, connection strings) there and move anything large into a bundled file, Netlify Blobs, or a fetch at runtime. There is no Netlify setting that raises this cap.
Only a function's own code and the modules it imports are bundled and deployed. A file the function opens from disk at runtime — fs.readFile/readFileSync on a template, a JSON data file, a WASM binary, a fixture — is not part of the bundle unless you declare it. This is a classic "works locally, ENOENT in production" trap: under netlify dev the function reads the file straight from your working tree, but the deployed function only contains what was bundled.
Declare runtime-read files with included_files in netlify.toml so they ship with the function:
[functions]
included_files = ["netlify/functions/templates/**"]
# or scope it to one function:
[functions."render-email"]
included_files = ["netlify/functions/templates/welcome.html"]
When the data is static, prefer importing it as a module (import data from "./data.json") so bundling is automatic; reach for included_files for files you must read from the filesystem at runtime.
| Resource | Limit |
|---|---|
| Synchronous timeout | 60 seconds |
| Background timeout | 15 minutes |
| Scheduled timeout | 30 seconds |
| Memory | 1024 MB default; configurable 1024–4096 MB (see Memory or vCPU) |
| Buffered payload | 6 MB |
| Streamed payload | 20 MB |
Frameworks with server-side capabilities (Astro, Next.js, Nuxt, SvelteKit, TanStack Start) typically generate their own serverless functions via adapters. You usually do not write raw Netlify Functions in these projects — the framework adapter handles server-side rendering and API routes. Write Netlify Functions directly when:
See the netlify-frameworks skill for adapter setup.
Guides writing Netlify Edge Functions for middleware, geolocation logic, request/response manipulation, auth checks, and A/B testing. Covers Deno runtime, config, path scoping, and edge vs serverless tradeoffs.
Guides developers in configuring, debugging, and optimizing server-side code running on Vercel — including Serverless Functions, Edge Functions, Fluid Compute, streaming, Cron Jobs, and runtime configuration.
Builds and deploys serverless applications on Cloudflare Workers using JavaScript, TypeScript, Python, or Rust. Useful for APIs, full-stack web apps, edge functions, background jobs, and real-time apps.
claude plugin install netlify-skills@claude-plugins-official