As AI agents move toward Vercel’s deepsec agentic coding, securing network access is a critical bottleneck. This update mirrors Cloudflare’s outbound security workers by providing a programmable egress layer. It prevents data exfiltration by ensuring agents only communicate with approved domains through a controlled, auditable intermediary.
You can now use matchers to apply proxying or credentials brokering (injecting secrets into requests) based on specific HTTP methods or paths. This allows for fine-grained control, such as only proxying POST requests to a specific API. These features are in beta for Pro and Enterprise plans via the @vercel/sandbox@beta SDK.
Vercel Sandbox firewall can now proxy and filter outbound requests. ▪︎ Control all agent traffic from your own proxy ▪︎ More granular credential injection ▪︎ Log and audit outgoing traffic https://t.co/CxemIM5C0h