Platforms

Deploying to Cloudflare Workers

Run RivetKit on Cloudflare Workers with the WebAssembly runtime.

Cloudflare Workers run RivetKit through the WebAssembly runtime. Use the public @rivetkit/rivetkit-wasm package, pass the bindings through setup({ wasm }), and use remote SQLite.

Steps

Prerequisites

Cloudflare account
wrangler configured for your account
A Rivet namespace from the Rivet Dashboard or a self-hosted Rivet Engine

Install Packages

npm install rivetkit @rivetkit/rivetkit-wasm
npm install --save-dev wrangler

Configure Wrangler

Set your Rivet connection values as Worker variables. The pool name must match the serverless runner configured in Rivet.

name = "rivetkit-cloudflare"
main = "src/index.ts"
compatibility_date = "2025-04-01"
compatibility_flags = ["nodejs_compat"]

[vars]
RIVET_ENDPOINT = "https://api.rivet.dev"
RIVET_NAMESPACE = "your-namespace"
RIVET_POOL = "cloudflare-workers"
RIVET_TOKEN = "sk_..."
RIVET_PUBLIC_ENDPOINT = "https://your-namespace:pk_...@api.rivet.dev"

Create the Worker

This example uses raw SQL to keep the runtime setup visible. When runtime: "wasm" is used, unset SQLite defaults to remote SQLite, and sqlite: "local" is rejected.

import { actor, setup } from "rivetkit";
import * as wasmBindings from "@rivetkit/rivetkit-wasm";
import wasmModule from "@rivetkit/rivetkit-wasm/rivetkit_wasm_bg.wasm";

interface Env {
  RIVET_ENDPOINT: string;
  RIVET_NAMESPACE: string;
  RIVET_POOL: string;
  RIVET_TOKEN: string;
  RIVET_PUBLIC_ENDPOINT: string;
}

interface SqliteDatabase {
  run(sql: string, params?: unknown[]): Promise<void>;
  query(sql: string, params?: unknown[]): Promise<{ rows: unknown[][] }>;
}

const rawSqlDatabaseProvider = {
  createClient: async () => ({
    execute: async () => [],
    close: async () => {},
  }),
  onMigrate: async () => {},
};

const counter = actor({
  db: rawSqlDatabaseProvider,
  actions: {
    increment: async (ctx, amount = 1) => {
      const db = ctx.sql as SqliteDatabase;
      await db.run(
        "CREATE TABLE IF NOT EXISTS counters (id INTEGER PRIMARY KEY, count INTEGER NOT NULL)",
      );
      await db.run(
        "INSERT INTO counters (id, count) VALUES (1, ?) ON CONFLICT(id) DO UPDATE SET count = count + excluded.count",
        [amount],
      );

      const result = await db.query("SELECT count FROM counters WHERE id = 1");
      return Number(result.rows[0]?.[0] ?? 0);
    },
  },
});

const use = { counter };
let registry: { handler(request: Request): Promise<Response> } | undefined;

function getRegistry(env: Env) {
  registry ??= setup({
    runtime: "wasm",
    sqlite: "remote",
    wasm: {
      bindings: wasmBindings,
      initInput: wasmModule,
    },
    use,
    endpoint: env.RIVET_ENDPOINT,
    namespace: env.RIVET_NAMESPACE,
    token: env.RIVET_TOKEN,
    envoy: {
      poolName: env.RIVET_POOL,
    },
    serverless: {
      publicEndpoint: env.RIVET_PUBLIC_ENDPOINT,
    },
  });

  return registry;
}

export default {
  async fetch(request: Request, env: Env): Promise<Response> {
    return await getRegistry(env).handler(request);
  },
};

Deploy

npx wrangler deploy

After deploy, set the Worker URL with the /api/rivet path as the serverless runner URL in Rivet.

Runtime Notes

Use runtime: "wasm" in setup(...) for Workers. You can also set RIVETKIT_RUNTIME=wasm in environments where the registry config does not set runtime.
Pass wasm: { bindings, initInput } explicitly from @rivetkit/rivetkit-wasm.
Use remote SQLite on Workers. Leaving SQLite unset with runtime: "wasm" selects remote SQLite automatically.
Keep RIVET_PUBLIC_ENDPOINT pointed at the client-facing Rivet endpoint. Register the Worker URL separately as the serverless runner URL.
Local Workers runtimes must support outbound WebSockets for the Rivet envoy connection.