Vercel has launched Security Review in public beta, a new capability inside Vercel Agent that sits alongside Code Review and Investigation. Once enabled, it scans for vulnerabilities on every push to a team's connected GitHub repositories.

Findings surface as pull request comments, in the Security dashboard, and through the CLI. With your permission, Vercel Agent can also validate findings and push fixes.

Security reviews are triggered by GitHub push events. Security reviews are triggered by GitHub push events. Security reviews are triggered by GitHub push events. Security reviews are triggered by GitHub push events.

Triggering and scope

Reviews are fired by GitHub push events, including direct pushes to a production branch. On monorepos, the scan is limited to the project's root directory.

Each review works off a diff rather than the full codebase: the push is compared against a base, and only changes are covered. For deployments, that base is the branch's previous deployment, so the agent evaluates the change against what is already running. Diff-only scanning keeps completion time in the seconds, which allows Security Review to run on every preview and production deployment.

Review methodology and policies

The agent does not limit itself to the changed lines. It reads the diff in the context of surrounding code, traces how the changed code can be reached, checks for existing protections, and rechecks every finding before reporting it. This is what lets it catch logic flaws such as missing authorization checks that pattern-matching scanners miss.

Teams can also write security policies in plain language and scope them to specific projects or the whole team. One example: requiring any endpoint that returns customer data to verify that the requester owns it. Policies can be managed from the Vercel CLI via vercel security policy.

Keeping findings deduplicated

New findings are compared against previously reported ones so the same problem is not flagged twice. Each finding is bound to the deployment where it first appeared, and tracked from there. If a vulnerable change from a pull request is merged into the production branch, it is flagged in production as well. When a later review no longer detects an issue in an environment, the dashboard marks it as no longer detected there.

Acting on findings

Every finding documents the vulnerability and its remediation, with links to the originating commit and deployment. Findings live in the Vulnerability view of the Security dashboard, where Vercel Agent can be asked to validate and commit fixes, and are also available via CLI for teams fixing with their own coding agent. Each new finding emits an audit log event, which Enterprise teams can forward to a SIEM through Audit Log Drains.

Fixes from Vercel Agent

Where a finding is fixable, the finding page offers Fix with Vercel Agent. The agent first confirms the vulnerability still exists, then applies the smallest possible change and updates the tests.

Take an invoice download endpoint whose lookup fails to verify which organization owns the invoice. Within seconds, Security Review reports that any signed-in user could fetch another organization's invoice by changing its ID. The agent's fix scopes the lookup to the requesting user's organization:

download-invoice.ts

const { organizationId } = await requireUser();

const invoice = await db.invoice.findUnique({

where: { id: invoiceId },

});

if (!invoice) return notFound();

return downloadInvoice(invoice);

Example authorization flaw and remediation in a fictional invoice endpoint.

External coding agents and triage

Findings can be pulled and fixed through the Vercel CLI by pointing your own coding agent (or another system) at it:

vercel security vulnerabilities

Dashboard findings can also be copied as Markdown or JSON, which makes them easy to paste into an agent conversation.

Motivation and availability

Frontier models are making vulnerabilities cheaper to find and exploit; last month researchers turned two decade-old flaws in PHP core into working remote code execution exploits with relatively little effort. At the same time, agents write a growing share of shipped code, each handling security review differently — or skipping it. Security Review is designed to make scanning every push on every project the default, governed by team policies.

The feature is available today in public beta for Pro and Enterprise teams. It is enabled in Vercel Agent Settings, either across all projects or a chosen subset.

Billing follows the Vercel Agent rate: model tokens at the provider's rate with no markup, plus $0.25 per million tokens. Security reviews are free until November 10, 2026; fixes made with Vercel Agent are billed at the rate above.

Initially, Security Review covers code pushed to GitHub. Next on the roadmap: deployments from the Vercel CLI and v0, repositories that do not deploy on Vercel, and suggested fixes applicable directly from the pull request.