Lizard

Lizard (lizard.build) runs AI agents in fast, low-cost cloud sandboxes and ships what they build. Each sandbox is a Firecracker microVM with its own kernel. Agents run AI-generated or untrusted code there, not on your machine. They get a shell, files, HTTPS URLs for open ports, and a built-in desktop and browser to open and test apps. What you get: - Sandboxes that boot in 228 ms median, the fastest in our 300-run test against E2B (264 ms), Daytona (819 ms) and Modal (929 ms) - Persistent Volumes that keep files after a sandbox stops - Ready templates: base for shell tasks, codex for coding agents, interpreter for Python - US East and EU West regions - Logs, CPU and memory metrics for every sandbox - Lizard SDK for TypeScript and Python, Lizard CLI, Lizard MCP and Lizard Skill for coding agents - Deploy the apps your agents build, with Managed Postgres, Managed Redis and S3-compatible storage, and no CI/CD to manage Pricing: Small (2 vCPU, 4 GB RAM) $0.0162/hour, Medium (4 vCPU, 8 GB RAM) $0.0324/hour, Large (8 vCPU, 16 GB RAM) $0.0648/hour. Billed per second while the sandbox runs. 2 TB of sandbox egress per account every 30 days is included. Lizard Pro costs $19/month and includes $19 of usage; past that, pay as you go. Start with a 7-day trial that includes $5 of usage. A card is required, and nothing is charged during the trial. Pro runs up to 100 sandboxes at once; Enterprise sets custom limits.