--- title: "Vercel Security Review Scans Every Push for Vulnerabilities" description: "Vercel Security Review, in public beta for Pro and Enterprise teams, has an AI agent check every GitHub push for flaws. Reviews are free until November 10." url: https://publishmyblog.com/website-tools/vercel-security-review-beta/ section: "Website Tools" author: "Saeed Shahriar" published: 2026-11-05T19:50:00.000Z updated: 2026-11-05T19:50:00.000Z --- # Vercel Security Review Scans Every Push for Vulnerabilities It reads only the code that changed, posts findings to a dashboard and pull requests, and can open a fix for you to approve. Vercel launched Vercel Security Review in public beta on November 3, and once you switch it on, an AI agent checks every push to your GitHub repo for vulnerabilities. It's open to Pro and Enterprise teams. **Reviews are free until November 10, 2026.** Do you deploy from GitHub? Check your Agent settings. ## How a review runs Every push to your GitHub repo starts a review. So direct pushes, and even force pushes to your production branch, get checked as well, says [Vercel's changelog](https://vercel.com/changelog/vercel-security-review-public-beta). **Each review covers only what changed.** The agent compares your push with the branch's previous deployment, so it checks the change against what's live. Vercel says it finishes in seconds. It does more than match patterns. Per [Vercel's announcement](https://vercel.com/blog/vercel-security-review), the agent: - reads the diff and the code around it - traces how the changed code can be reached - looks for protections already in place - rechecks each finding before it reports it Vercel says that's how it catches logic flaws. Its own example is an invoice download route with a missing authorization check, where any signed-in user could grab another organization's invoice just by changing the ID in the request. ## Where findings show up Every finding lands in your **Security dashboard** with the project, branch and commit, a summary, a suggested fix and an AI-assigned severity. Private and internal repos also get a GitHub check on the commit and a pull request comment. On public repos, your findings stay in the dashboard. Then you've got two ways to fix one: 1. Click Fix with Vercel Agent. It confirms the flaw still exists, makes the smallest change that fixes it, updates the tests and opens a pull request. Nothing merges until you approve. 2. Pull findings with `vercel security vulnerabilities` (Vercel CLI 65.10.0 or later), or copy them as Markdown or JSON for your own coding agent. You can also write security policies in plain language, for your whole team or one project. Vercel's sample: any endpoint that returns customer data must check the requester owns it. ## What it costs after November 10 From November 10, **you pay the Vercel Agent rate**: the model provider's token price with no markup, plus $0.25 per million tokens. Agent fixes cost the same. So your bill grows with every push to your GitHub repo and with how much code changes. Pushing dozens of small commits a day? Then you'll pay more. ## Why it matters Vercel points to two shifts: AI models that make flaws easier to exploit, and agents writing more of the code. Its example: last month, researchers turned two old flaws in PHP core into working exploits. That reads to me like [the PHP2Shell flaws](https://publishmyblog.com/website-tools/php2shell-critical-php-flaws/) PHP patched on October 21. Still, I think **the second shift is the one that hits small teams.** If Claude Code, Cursor or v0 writes half the routes in your app, who checks the authorization on each one? In most small shops I see, nobody does. That said, read the limits Vercel spells out before you trust it: - **It isn't a security audit.** - It reads source code only, with no runtime or penetration testing. - For now it only checks every push to your GitHub repo; Vercel CLI and v0 deployments, and repos that don't deploy on Vercel, come next. Anyway, my take is simple. Turn it on while it's free. Push your real code. You'll see how noisy it is. Had keys leak before, like in the [Northflank breach of customers' AI API keys](https://publishmyblog.com/website-tools/northflank-breach-ai-api-keys/)? Write a secrets policy first. ## Common Questions **Is Vercel Security Review free to use on the Hobby plan?** No. Vercel Security Review checks every push to your GitHub repo only on Vercel Pro and Enterprise teams, and reviews are free until November 10, 2026, then billed at the Vercel Agent rate. **Does Vercel Security Review replace a penetration test or security audit for my web app?** It doesn't. Vercel says Vercel Security Review reads source code changes only and runs no runtime or penetration tests, so it can find vulnerabilities without proving an application has none. **Can I fix Vercel Security Review findings with Claude Code or Cursor instead of Vercel Agent?** Yes. Every finding in your Security dashboard can go to your own coding agent: the `vercel security vulnerabilities` command in Vercel CLI 65.10.0 or later exports findings, or you can copy each one as Markdown or JSON. ## Sources - [Vercel Security Review now in public beta](https://vercel.com/changelog/vercel-security-review-public-beta), Vercel - [Introducing Vercel Security Review](https://vercel.com/blog/vercel-security-review), Vercel