Agent Runtime & Sandbox — Market

Verified claims and product-axis read for Agent Runtime & Sandbox. Every fact below is sourced; every product judgment traces back to underlying signals.

Verified facts

• Modal raised ~$355M at a ~$4.65B valuation in May 2026 for its serverless AI/ML compute platform. ↗ (financial)
• Agent runtime/sandbox emerged as a distinct, well-funded 2026 category providing the execution substrate beneath agent frameworks. ↗ (other)
• Sandboxes isolate agent-run code from the host — a security necessity as agents execute untrusted, model-generated code. ↗ (other)
• E2B and Modal anchor the code-execution side of agent runtimes; Browserbase anchors the browser-automation side. ↗ (other)
• Demand for agent runtimes rose sharply with coding agents (Devin-style) that must run and test their own generated code. ↗ (other)
• The agent-sandbox layer is viewed as a likely 2026 acquisition target as hyperscalers move to own the agent execution substrate. ↗ (other)
• E2B operates open-source code-execution sandboxes for AI agents and surpassed 1B+ sandboxes created. ↗ _(technical_spec)_
• Daytona provides secure, ephemeral development environments increasingly used as agent sandboxes. ↗ _(technical_spec)_
• Browserbase offers headless-browser infrastructure that lets AI agents drive web browsers at scale. ↗ _(technical_spec)_
• Runloop provides managed runtimes and sandboxes purpose-built for coding agents. ↗ _(technical_spec)_

See the Products and Strategy modules for the full product list and forward-looking judgment.