Function AppComing soon

Click Create function on the Function App card of the project's Services tab. Give it a name, pick a runtime, write the handler and click Deploy (or deploy a file from your terminal, below). We build it and it is live at https://<name>-<project>.<apps domain> in about a minute.

1. Every request to the address, with any method and path, calls your handler.
2. The project's variables are set for it: DATABASE_URL, REDIS_URL, the S3_* storage keys, AUTH_*, EMAIL_*, JOBS_* and your custom variables, as on the Environment tab. They are read each time the function starts, so new services and new passwords are picked up.
3. After 10 minutes without requests it stops. The next request starts it again in about a second: it is held while the function starts and then sent on, with its method and body (a 307 redirect to the same address).
4. You pay for requests and the time they take, not for the hours it waits.

Node.js 24. Put the handler in index.ts (or index.js) as the default export. It gets a standard web Request and returns a Response; a string or an object is sent for you, as text or JSON. TypeScript runs as is: types are removed, so use plain syntax (no enum or namespace).

import pg from "pg";

// Created once per start, and reused by every request.
const pool = new pg.Pool({ connectionString: process.env.DATABASE_URL, max: 5 });

export default async function handler(request: Request): Promise<Response> {
  const url = new URL(request.url);
  if (request.method === "POST") {
    const { title } = await request.json();
    const { rows } = await pool.query("insert into todos (title) values ($1) returning *", [title]);
    return Response.json(rows[0], { status: 201 });
  }
  const { rows } = await pool.query("select * from todos order by id desc limit $1", [url.searchParams.get("limit") ?? 20]);
  return Response.json(rows);
}

{
  "dependencies": {
    "pg": "^8.13.0"
  }
}

Dependencies are installed with npm install on every deploy. The handler is an ES module unless package.json sets another type.

Python 3.13. Define handler(request) in main.py, plain or async. request has method, path, query (a dict), headers (lowercase keys), body (bytes), text and json(). Return a dict or list (sent as JSON), a string, bytes, None (204), or a tuple of the body, the status and a dict of headers.

import os
import psycopg

def handler(request):
    if request.method == "POST":
        title = request.json()["title"]
        with psycopg.connect(os.environ["DATABASE_URL"]) as db:
            row = db.execute("insert into todos (title) values (%s) returning id", [title]).fetchone()
        return {"id": row[0], "title": title}, 201
    return {"hello": request.query.get("name", "world")}

psycopg[binary]==3.2.3

curl "https://api-my-project.example.app/todos?limit=5"
curl -X POST https://api-my-project.example.app/todos -H "content-type: application/json" -d '{"title":"Ship it"}'

• The address is public. Check a token or a signature in the handler for anything private, for example a header your app sends with a secret from its own variables.
• Answer CORS yourself when browsers call it from another site: return the access-control-allow-* headers, and answer OPTIONS.
• An uncaught error answers 500 and is written to the Logs tab. A request that takes longer than the plan's timeout answers 504.
• Clients that do not follow redirects (curl without -L) may see the 307 on the first request after the function slept.

You can also deploy a file without opening the dashboard. Create function shows these steps with your project id filled in, under Deploy from terminal.

export default {
  fetch: (request: Request) => new Response("Hello world"),
};

export CLOUDHIKE_API_KEY="icu_..."

curl -X PUT "https://cloudhike.dev/api/v1/projects/<project id>/functions/hello?file=hello.ts" \
  -H "Authorization: Bearer $CLOUDHIKE_API_KEY" \
  --data-binary @hello.ts

• The last part of the address is the function's name. Run the command again to deploy a change; use another name to add another function to the project.
• The file's extension picks the runtime: .ts, .js or .mjs for Node.js, .py for Python. A function's runtime cannot change once it exists.
• Add &environment=development to deploy to another environment, and &size=small to choose a size.
• GET the same address for the function's status and url. GET .../functions lists the project's functions.
• A JSON body works too: {"runtime": "node", "code": "...", "dependencies": "..."} with content-type: application/json.

Edit the code on the function's Code & test tab and click Save and deploy; the running version keeps answering until the new one replaces it. The Test panel below it sends a request (any method, path, headers and body) to the deployed function and shows the status, time and response. Test requests are billed like any other. The row of names under the function's title moves between the project's functions, and New function adds one. Settings changes the size and adds variables of its own, which win over the project's. Logs shows the last 200 lines it wrote. AI editors connected through the MCP server can create and update functions with the deploy_function tool.

• GB-seconds are the time requests take, from arriving to the last byte of the answer, at the function's memory size: one second at 256 MB is 0.25 GB-seconds.
• Usage is counted every few minutes, so a Free function can go a little past its month before it stops. Upgrading to Pro starts it again.
• Requests and compute are charged from your credits every hour, like email and background jobs. The function's page shows this month's use.
• The function's disk is read-only; /tmp can be written, is kept in memory and is emptied when it stops. Keep data in Postgres, Redis or Storage.