# spawn > Firecracker microVM sandboxes for AI agents. Each gets its own kernel, memory and disk. Run code, open a terminal or IDE, then destroy it in one API call. spawn is in private beta. The HTTP API documented below is live; official Python and TypeScript SDKs, browser automation and custom images are not available yet. ## For AI agents - Every docs link below is Markdown. Open only the pages the task needs; all of them are also in one file at https://runspawn.com/llms-full.txt. - The API's base URL is https://dev-api.renderops.com. Every endpoint except `/health` needs `Authorization: Bearer `. Keys start with `ro_live_` and are created at https://runspawn.com/dashboard/keys. - Request and response bodies are JSON. Treat the HTTP status as the source of truth for errors; read the JSON `code` only when the body is JSON. ## Product - [Use cases](https://runspawn.com/use-cases): Agent swarms on a repo, security review on every PR, RL training environments, and scripting sandboxes from the CLI. ## Getting started - [Introduction](https://runspawn.com/docs.md): spawn gives every AI agent its own Firecracker microVM sandbox. How the HTTP API creates, runs, pauses and destroys sandboxes. - [Quickstart](https://runspawn.com/docs/quickstart.md): Create a sandbox, run Python in it, run a shell command and destroy it, in five HTTP requests with cURL, Python or JavaScript. - [Authentication](https://runspawn.com/docs/authentication.md): Authenticate spawn API requests with a bearer API key, what happens when a key is rejected, and the single-use tokens for terminals and previews. ## CLI - [The spawn CLI](https://runspawn.com/docs/cli.md): Run Claude Code in a sandbox from your terminal: sign in with spawn login, upload a project with spawn claude, bring its changes back with spawn pull. - [Claude sessions](https://runspawn.com/docs/cli/sessions.md): What spawn uploads and leaves out, how Claude gets your key, resuming, sizes, detaching, pausing, pulling changes, pull requests and background agents. - [CLI reference](https://runspawn.com/docs/cli/reference.md): Every spawn command and flag, the settings files and keys, environment variables, local state and exit codes. ## Sandboxes - [Create a sandbox](https://runspawn.com/docs/sandboxes.md): Create a Firecracker microVM sandbox with POST /session: sizes, images, network access, idle timeout, maximum lifetime and the response. - [Run code and commands](https://runspawn.com/docs/run.md): Run Python, Node or Bash code and shell commands inside a spawn sandbox, with warm runtimes, timeouts and how failures are reported. - [Pause, resume, destroy](https://runspawn.com/docs/lifecycle.md): Pause a sandbox to free its CPU and memory, resume it with its disk intact, or destroy it. Plus automatic idle pause and cleanup. ## Open it up - [Terminals](https://runspawn.com/docs/terminals.md): Attach a live, interactive shell to a running sandbox over WebSocket with a single-use token, and the messages it sends and receives. - [Browser IDE](https://runspawn.com/docs/ide.md): Open VS Code in the browser inside a running spawn sandbox with a single-use signed link. - [Port previews](https://runspawn.com/docs/previews.md): Create signed, private preview links to web servers running inside a sandbox, valid from one second to 24 hours. ## REST API - [Endpoints](https://runspawn.com/docs/api.md): Every public spawn API endpoint on one page: sandboxes, code execution, lifecycle, terminals, browser IDE and port previews. - [Requests and errors](https://runspawn.com/docs/errors.md): The spawn API’s base URL, headers, request IDs, JSON error format and every HTTP status code it returns. - [Limits](https://runspawn.com/docs/limits.md): Sandbox sizes, images, languages, token lifetimes, preview expiry and the other fixed limits of the spawn API during the private beta. ## Optional - [All docs in one file](https://runspawn.com/llms-full.txt): every page above as Markdown, in reading order. - [Privacy Policy](https://runspawn.com/privacy) - [Terms of Service](https://runspawn.com/terms)