AI coding sandboxes, self-hosted

Spin up AI dev sandboxes
in seconds, not hours.

Isolated VS Code workspaces with Claude Code pre-installed, metered through your own LLM gateway. Your provider keys never leave the server.

sandbox — terminal

$ claude "add dark mode to the settings page"

✓ Authenticated via LLM gateway (virtual key)

✓ Reading src/pages/settings.tsx ...

Editing 3 files — settings.tsx, theme.ts, globals.css

2,431 tokens used · within key limit

Everything between your team and the model

A control plane for AI-assisted development: sandboxes, gateway, keys and analytics in one place.

Isolated sandboxes

Every session runs in its own container with CPU, memory and lifetime limits. Nothing leaks between workspaces.

VS Code in the browser

One click spawns a full VS Code workspace — terminal, extensions, file tree — served straight from the sandbox.

Claude Code pre-installed

Each sandbox boots with the Claude Code CLI ready in the terminal, already authenticated against the gateway.

LLM Gateway

Bring your own provider — Azure OpenAI, OpenAI, Anthropic or OpenRouter. One endpoint, your keys stay server-side.

Virtual API keys

Issue revocable keys with per-key token limits. The real provider key is never exposed to any sandbox.

Usage analytics

Token usage tracked per request, per key and per model — live charts in the dashboard plus exportable PNG reports.

How it works

From zero to an AI-powered workspace in three steps.

0101

Configure the gateway

An admin points the LLM gateway at your provider — endpoint, model and API key, stored server-side only.

0202

Spawn a sandbox

Hit "New session" and get an isolated VS Code workspace in seconds, with a fresh virtual key injected automatically.

0303

Build with Claude Code

Open the terminal, run claude, and ship. Every token is metered against your key — watch it live in Usage.

Ready to spawn your first sandbox?

Sign in with the account your admin provisioned and you're one click away from a workspace.

Sign in to start