Solari

Teams building AI agents need isolated environments to browse the web, run code, or operate software, and as agent usage grows these environments must start quickly, stay contained, and run in parallel. Today this often means stitching together separate infrastructure for browsers, sandboxes, and virtual machines from different providers. Benchmarks cited on the page show competing browser and sandbox providers taking anywhere from hundreds of milliseconds to several seconds to create, connect, and release environments, which adds latency when agents run at scale. Without a unified layer, teams must manage authentication, session handling, and scaling separately across each environment type.

Solari offers one API that gives agents access to three environment types: cloud browsers, AI sandboxes, and computer desktops. According to the page, browsers spin up in 8ms, sandboxes in 90ms, and desktops resume from snapshot in 0.78ms, and the site presents latency benchmarks against providers like Steel, Browserbase, Hyperbrowser, Anchorbrowser, E2B, Modal, Daytona, and CodeSandbox showing faster end-to-end performance. Browsers include stealth mode, residential proxies, automatic CAPTCHA solving, and persistent profiles for staying logged in. Sandboxes run untrusted AI-generated code on hardware-isolated microVMs. Machines can pause automatically when idle, resume later, and be snapshotted, rewound, or forked into parallel copies, and custom templates let users pre-install packages, tools, and configuration via an image builder. The platform supports TypeScript, Python, Go, Rust, and C++, and Team and Enterprise customers can deploy within their own VPC.