Cloudflare Sandbox

Cloudflare Sandbox uses @cloudflare/sandbox to provide a container-backed Linux environment to a Flue application deployed on Cloudflare. This integration is platform-native: it is not an adapter module for a Node-target application.

Quickstart

Add container-backed Linux sandbox capability to an existing Flue project with the Cloudflare Sandbox blueprint. Run the following command in your terminal or coding agent of choice:

flue add sandbox cloudflare

Overview

Cloudflare Sandbox is a Cloudflare target integration rather than a generated adapter. In a Cloudflare-targeted project, the blueprint installs @cloudflare/sandbox; a workflow obtains the bound Durable Object with getSandbox(...), wraps it with Flue’s cloudflareSandbox(...), and passes that sandbox factory to a created agent.

import { createAgent, type FlueContext, type WorkflowRouteHandler } from '@flue/runtime';
import { cloudflareSandbox } from '@flue/runtime/cloudflare';
import { getSandbox } from '@cloudflare/sandbox';

export const route: WorkflowRouteHandler = async (_c, next) => next();

export async function run({ init, id, env, payload }: FlueContext<{ message: string }>) {
  const sandbox = cloudflareSandbox(getSandbox(env.Sandbox, id));
  const agent = createAgent(() => ({ sandbox, model: 'anthropic/claude-opus-4-7' }));
  const harness = await init(agent);
  const session = await harness.session();

  return await session.prompt(payload.message);
}

The blueprint also exports Sandbox from <source-root>/cloudflare.ts, adds its Durable Object binding, a new migration entry, and its container declaration to wrangler.jsonc, and creates a project-root Dockerfile whose image tag matches the installed package version. The resulting workflow runs agent shell and file operations in the container-backed sandbox identified by the workflow run id. Cloudflare’s direct delete API does not expose recursive or force controls, so cloudflareSandbox() rejects either option before mutation. A Node-targeted project must migrate to the Cloudflare target before using this integration.

Configure

Requirement	Purpose
Cloudflare target	Required — Runs the platform-native sandbox integration.
@cloudflare/sandbox package	Required — Provides the Sandbox Durable Object and RPC client.
Container image	Required — Defines the Linux filesystem and command environment.
Durable Object/container binding	Required on Cloudflare — Exposes the sandbox through Wrangler platform configuration; it is not an environment variable.
Stable sandbox identity and retention policy	Required — Controls lifecycle and reuse for the application.
Environment-variable credentials	Not required — The platform integration uses Cloudflare bindings and deployment configuration instead.

Cloudflare Sandbox requires a Worker deployment, Durable Object/container configuration, and a container image. Add the dependency to a Cloudflare-targeted project and export its Durable Object class from your Cloudflare deployment module:

// <source-root>/cloudflare.ts
export { Sandbox } from '@cloudflare/sandbox';

Declare the sandbox binding in Wrangler configuration, then wrap the RPC stub returned by getSandbox(...) with cloudflareSandbox(...) and pass it to an agent:

import { getSandbox } from '@cloudflare/sandbox';
import { createAgent } from '@flue/runtime';
import { cloudflareSandbox } from '@flue/runtime/cloudflare';

type Env = { Sandbox: DurableObjectNamespace };

export default createAgent<unknown, Env>(({ id, env }) => ({
  model: 'anthropic/claude-sonnet-4-6',
  sandbox: cloudflareSandbox(getSandbox(env.Sandbox, id)),
  cwd: '/workspace',
}));

Choose this integration when

Use Cloudflare Sandbox when an agent on Cloudflare needs git, package installation, native binaries, or other Linux tooling. Prefer Cloudflare Shell instead when a durable workspace with Workspace-oriented operations is sufficient and a Linux toolchain is unnecessary.

Treat network egress, mounted data, credentials, and side effects as application security decisions. See Sandboxes and Deploy on Cloudflare.