---
title: Ibby docs
description: Run code in isolated sandboxes from your app, with a TypeScript SDK, a Python SDK, a command line tool or plain HTTP.
---

Ibby runs your code in isolated sandboxes. A sandbox has no network, holds no secrets and stops on its own when its time is up. You create one, put files in it, run commands, read the results and remove it.

You can reach Ibby four ways. All four do the same things.

| Way | Install | Page |
| --- | --- | --- |
| TypeScript SDK | `npm install @ibby-ai/sdk` | [TypeScript SDK](/sdks/typescript) |
| Python SDK | `pip install ibby` | [Python SDK](/sdks/python) |
| Command line | `npm install --global @ibby-ai/cli` | [The ibby command](/cli) |
| HTTP | Nothing to install | [API reference](/api) |

## Where to start

- New here? Follow the [quickstart](/quickstart). It takes a few minutes and needs only your API key.
- Want the ideas first? Read [concepts](/concepts): lanes, keys, sandboxes, the store folder, images, jobs, retries and errors.
- Want to see what your runs cost? [Usage and cost](/usage) lists them, with estimates.
- Looking up one call? The [API reference](/api) has a page for every operation.

## Keep keys and client data out

Never paste an API key, a secret or any client detail into these docs, their search or their assistant. Use synthetic data in sandboxes.