Willison and Prime Radiant Release External Sandbox Mode
Understand this story
Simon Willison, in collaboration with AI research lab Prime Radiant, released an external sandbox mode for coding agents: the code executed by the agent runs inside an isolated sandbox, while the agent itself stays outside the sandbox. To demonstrate this mode, the two parties released three sandbox server implementations, of which external-sandbox-v86 runs a Node.js server and then uses v86 to run Alpine Linux in WebAssembly. These sandboxes also include an optional HTTP/HTTPS allowlist proxy to restrict which hosts the agent code can communicate with, and provide a credential substitution mechanism that lets the agent make authenticated API calls without knowing the keys.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Simon WillisonExternal sandboxes for coding agents
The post introduces the external sandbox pattern, in which code executed by an agent runs inside a dedicated sandbox while the agent itself lives outside it, aimed at making it safer to try local models against mistakes or prompt injection. Three sandbox server implementations were released with the same localhost API: external-sandbox-v86 (Node.js server running Alpine Linux in WebAssembly via v86), external-sandbox-container (Python web server running Alpine Linux in an Apple Containers instance), and external-sandbox-quicksand (Python web server using Microsoft's Quicksand library, bundling QEMU and Alpine Linux). Pi, Claude Code, or Codex can be configured to send API calls to the sandbox for code execution, and the sandboxes include an optional HTTP/HTTPS allow-list proxy plus a credential replacement mechanism for authenticated API calls without sharing a secret with the agent.
Activity for this story
Not enough continuous observations to show a trend yet.