Welcome to LobsterCage

Secure, isolated cloud sandboxes for AI agents.

LobsterCage provides containerized environments where AI agents can run code safely — with no inbound ports, persistent storage, and automatic lifecycle management.

Why LobsterCage?

• Fully isolated — Each cage is an isolated cloud container with no inbound network access. Agents can reach the internet, but nothing can reach them uninvited.
• Hibernate when idle — Cages automatically hibernate after inactivity (default: 30 minutes) and wake on demand. You only pay for active compute.
• Persistent storage — Persistent storage survives hibernation and restarts. Your agent’s data stays put.
• Webhook gateway — Receive Telegram messages via webhook. Requests are buffered and delivered when the cage wakes.
• SSH & tunnels — Connect via SSH, port-forward, or browser terminal. Binary-safe WebSocket relay for SSH; no SSM corruption.
• Simple CLI — Deploy, connect, and manage cages with the lobster command.

How it works

1. Deploy a cage from the CLI or dashboard
2. Connect via SSH, browser shell, or API
3. Cage hibernates when idle to save costs
4. Incoming webhooks wake the cage automatically
5. Pay only for active compute time

Deployment patterns

• Why multiple agents? — Why dedicated, purpose-built agents outperform a single overloaded bot
• Single agent setup — Deploy one agent with hibernation and auto-wake
• Multi-agent direct — Multiple agents, each on its own messaging channel
• Orchestrator pattern — An agent that creates and manages other agents

Get started

Follow the Quickstart to deploy your first cage in under 5 minutes.