Vercel Makes Sandbox Egress Firewall Free Across Every Plan.

Vercel is removing the price barrier from one of its more important safeguards for running untrusted code. The company’s Sandbox Egress Firewall, previously a paid add-on, is now included on every Vercel plan at no additional cost.

The company announced the change on LinkedIn, arguing that compute isolation alone is not enough to contain potentially hostile code. A sandboxed process may be restricted at the runtime level, but outbound network access can still create opportunities to leak data, contact external infrastructure, or probe services that should remain inaccessible.

Making the firewall free broadens access to network-level containment at a time when more developers are allowing AI agents, code generators, and user-controlled workflows to execute code dynamically.

Traditional sandboxing focuses on separating a process from the host system. It may limit access to files, memory, system calls, and other workloads. That remains necessary, but it does not automatically control where the process can connect over the network.

An egress firewall addresses that second layer. It governs outbound traffic from the sandbox, helping developers restrict the destinations that executed code can reach. This matters when an application handles generated scripts, third-party packages, uploaded programs, or commands produced by an AI model.

Without a network boundary, compromised or deliberately malicious code could attempt to:

• Send sensitive information to an external server
• Reach unauthorized APIs or internal services
• Download additional payloads or dependencies
• Communicate with infrastructure controlled by an attacker

Vercel’s position is summarized by the title of its related engineering article, “A sandbox without a network boundary is only half a sandbox”. The company connected the pricing change to recent security research, although the announcement did not provide detailed configuration instructions or identify specific default firewall policies.

The update complements Vercel’s broader work on Sandbox infrastructure. It follows the platform’s move to run Vercel Sandbox on Vercel Managed Images, which shifted responsibility for underlying execution images toward Vercel’s managed environment.

The immediate difference is financial. Teams no longer need to purchase the firewall as an add-on to place a network boundary around untrusted Sandbox workloads. Because Vercel says the feature is free on every plan, the change also applies to individuals and smaller teams that may not have budgeted for a separate security product.

That is particularly relevant for builders creating AI coding tools, browser-based development environments, automated testing systems, or applications that accept and execute user-provided code. These products often need internet access for legitimate tasks, such as calling an API or retrieving a dependency, while still requiring controls that prevent arbitrary outbound connections.

The firewall should not be treated as a complete security strategy. Runtime isolation, secret management, dependency controls, logging, and narrowly scoped permissions remain important. Network restrictions add another boundary, reducing the routes available to untrusted code if another defensive layer fails.

Vercel has not announced changes to Sandbox compute pricing, usage allowances, or plan limits as part of this update. The stated change concerns the cost and plan availability of the egress firewall itself. Developers evaluating the feature should therefore review current Sandbox documentation and their existing usage limits before moving production workloads.

What is Vercel Sandbox Egress Firewall?

Vercel Sandbox Egress Firewall is a network security layer for code running inside Vercel Sandbox. It controls outbound network access, adding network-boundary isolation on top of the runtime and compute isolation provided by the sandbox.

How much does Vercel Sandbox Egress Firewall cost?

Vercel says the firewall is now free on every plan. It was previously offered as a paid add-on, but the announcement does not indicate that other Sandbox compute charges or usage limits have changed.

Who should use the firewall?

It is relevant to developers running untrusted or dynamically generated code, including AI agent workflows, code execution tools, automated testing systems, and applications that process user-submitted programs. It adds a safeguard against unauthorized outbound network activity but does not replace other security controls.