Which sandbox is best for Cursor?
It depends on your Cursor plan and on who should run the machine. For an agent that starts its own sandboxes, Runtime connects through one MCP entry that holds no secret. Of the services priced below, it costs the least: $16.56 for fifteen hundred tasks, compared with $33.01 on Cloudflare Sandbox and $82.80 on E2B. Keep your own machine for quick, trusted edits. Pick Cursor's Cloud Agents to have Cursor run the machine, or Self-Hosted Machines for a worker you host.
Where can Cursor's agent run code?
Cursor gives its agent three places to work. These are the places its docs described on 25 October 2026:
| Where | Who runs the machine | What it needs |
|---|---|---|
Your machine, in the editor or agent CLI |
You | Nothing; commands run on your laptop |
| Cursor's Cloud Agents | Cursor | A Cursor plan with Cloud Agents |
| Self-Hosted Machines | You, or a sandbox provider Cursor lists | A worker; shared Team Pools need Enterprise |
For Self-Hosted Machines, Cursor's integrations page lists partner guides for AWS Lambda, Cloudflare, Namespace, Modal, Daytona, E2B, Vercel, Tensorlake, Coder and Superserve. A worker opens "an outbound HTTPS connection to Cursor, and Cursor sends agent tool calls over that connection". It needs no inbound ports.
A fourth way is an MCP server in mcp.json that lets the agent start a
sandbox when it needs one.
What should Cursor's agent get from a sandbox?
- Setup an agent can do alone. One approval. Nothing secret goes in
.cursor/mcp.json, so the project file is safe to commit. - A machine per idea. A fresh machine for every experiment, and a way to copy one with its memory. Two fixes can then be tried side by side from the same state.
- Servers you can open. An app the agent starts should get a private HTTPS address that you open in your browser.
- Your laptop left alone. A runaway
npm install, a fork bomb or a test that wipes a directory runs on a machine made for it. - A cost that follows use. When CPU is billed on what the code actually uses, an agent that thinks for minutes between commands pays little for the wait.
What does a day of Cursor agent work cost?
Fifteen hundred agent tasks a month, each twenty minutes on a 2 vCPU, 4 GiB sandbox and busy for 150 CPU-seconds, at each provider's published rates:
| Provider | Isolation | A month of tasks | Runtime costs less by |
|---|---|---|---|
| Runtime Cloud | Firecracker microVM | $16.56 | |
| Cloudflare Sandbox | A container in its own VM | $33.01 | 50% |
| Vercel Sandbox | Firecracker microVM | $50.40 | 67% |
| E2B | Firecracker microVM | $82.80 | 80% |
| Daytona | Containers by default | $82.80 | 80% |
| Modal | gVisor, a shared kernel | $118.98 | 86% |
| AWS Lambda MicroVMs | Firecracker microVM | $126.10 | 87% |
When might another sandbox fit better?
- Cloud Agents started from Cursor's own interface. Self-Hosted Machines run those agents on a worker you host; shared Team Pools need Cursor Enterprise. Cursor's partner guides cover the providers above.
Runtime Cloud
Runtime gives Cursor's agent, in the editor or the agent CLI, a fresh
Firecracker microVM for every experiment. The agent reaches it through one MCP
entry that holds no secret. The agent keeps editing on your machine, while
builds, test suites and servers run in a machine with its own Linux kernel.
- The first Runtime tool call prints a link and a code. You approve Connect agent once, and the agent then has every tool.
runtime_sandboxes_createreturns a running microVM.runtime_sandboxes_forkcopies one with its memory.- An app the agent starts gets a private HTTPS preview address.
- A sandbox waiting for the agent's next step costs $0.03125 an hour for 2 vCPU and 4 GiB. It pauses itself after 300 seconds with nothing happening, so an idle sandbox the agent forgot about costs only storage.
The fifteen hundred tasks above cost $16.56 on Runtime, 50% less than Cloudflare Sandbox and 80% less than E2B, at rates checked 25 October 2026.
A run on Runtime, recorded 1 October 2026
Add the server once, for every project:
JSON{ "mcpServers": { "runtime": { "command": "npx", "args": ["-y", "withruntime", "mcp"] } }}Or run the agent CLI itself inside a sandbox, for a job nobody watches:
TypeScriptimport { Sandbox } from "withruntime";await using sbx = await Sandbox.create({ timeoutSeconds: 3600 });await sbx.exec("curl -fsSL https://cursor.com/install | bash", { check: true, timeoutMs: 600_000 });console.log((await sbx.exec("~/.local/bin/agent --version")).stdout);On 1 October 2026, that install ran in a fresh production sandbox:
textcurl -fsSL https://cursor.com/install | bash exit 0~/.local/bin/agent --version 2026.10.01-e373342Cursor's installer ran unchanged on Ubuntu 24.04 and put agent in
~/.local/bin. Print mode, --approve-mcps and a prepared sandbox for CI are
in Cursor CLI in a sandbox. The editor setup is in
add Runtime to Cursor.
The agent's first sandboxes run on free usage. The first tool call signs it
in, or you can run npx withruntime login yourself. Code that already calls
another provider moves with the migration guide.
Sources
Checked 25 October 2026.
- Cursor self-hosted machines: integrations: the partner guides, the Enterprise and service-account requirements, and how a worker connects
- Cursor self-hosted machines: My Machines and Team Pools, and no inbound ports
- Each provider's published rates, checked 25 October 2026, as the pricing guide lists them
- The recorded run: Cursor's installer and
agent2026.10.01-e373342, in a production sandbox on 1 October 2026