On this page
Vercel change
Vercel Security Review (public beta, November 3, 2026): how to turn it on, what each push review covers, and what it costs after November 10
Short answer
Vercel Security Review, in public beta for Pro and Enterprise teams since November 3, 2026, reviews the code changed by every push to a connected GitHub repository and reports findings in the Vulnerabilities dashboard, as a GitHub check, and as pull request comments on private repositories. Reviews are free until November 10, 2026, then billed at the Vercel Agent rate: provider tokens at cost plus $0.25 per million tokens.
Who it affects
Pro and Enterprise teams that deploy from GitHub, and coding agents that triage findings. A team Owner turns it on in the Agent settings. Vercel says it analyzes source code only: it is not a security audit and does no runtime or penetration testing, so a passing review does not mean an app has no vulnerabilities.
What a review covers
Reviews start from push events, not pull requests, so direct pushes and force pushes to the production branch are reviewed too, and every project in one push is reviewed together. Each review compares the source with a base, the previous deployment of the branch, so it covers what changed; a project's first review reads the full source. In a monorepo each project is reviewed from its own root directory.
A finding names the project, environment, branch and commit, with a summary, details, a suggested fix and a severity of Critical, High, Medium or Low. Vercel says the severity is an AI-generated estimate, not a fixed rubric. A finding is Open until you fix or dismiss it; Closed means a later review no longer detects it, which Vercel says does not confirm the issue is fixed.
Triage findings from the terminal or an agent
Vercel CLI 65.10.0 or later lists and triages findings with vercel security vulnerabilities; list is the default and shows open findings, 20 at a time. inspect, dismiss and reopen take a reference such as my-app/12. dismiss needs --reason false-positive or accepted-risk in a non-interactive run, or it fails. The command needs a team scope (vercel switch) and fails if Security Review is not enabled for the team.
Vercel warns that findings include source code and how to exploit it, so keep this output out of shared or public CI logs.
vercel security vulnerabilities --severity critical,high --json
vercel security vulnerabilities inspect my-app/12
vercel security vulnerabilities dismiss my-app/12 --reason accepted-risk --message "Internal tool behind SSO"
vercel security vulnerabilities reopen my-app/12What it costs
Reviews are free until November 10, 2026. After that, each review is billed at the Vercel Agent rate: provider inference at the underlying token rate with no markup, plus the Vercel Token Rate of $0.25 per million input, cached and output tokens. Reviews run per push, so cost grows with how often you push and how much code changes. Fixes made with Vercel Agent are billed at the same rate.
Turn it on in five steps
- 1As a team Owner, open Settings, then Agent. In the Security Reviews section, turn on Review Deployments Automatically.
- 2Under Projects to Review, keep All projects or pick up to 100 projects, then click Save. Only GitHub-connected repositories are reviewed.
- 3To use Fix with Vercel Agent, also turn on Enable Chat on the same page. The agent opens a pull request against the finding's branch, and nothing merges until you approve it.
- 4Optional: write a policy in plain language, up to 32,000 characters, in the team Agent settings or a project's Security settings, for example which routes need sign-in. Vercel treats it as context, and a policy that tells reviews to skip something can make them miss a real issue.
- 5Before November 10, 2026, check how often your team and agents push, since every push to a connected repository starts a billed review after that date.
Where Security Review surprises teams
From Vercel's changelog entry and the Security Review and vercel security docs, read on 2026-11-05.
Pull request comments are wider than the dashboard
On a private or internal repository, findings are posted as pull request comments that anyone with access to the repository can read, while the dashboard hides them from the Billing and Viewer roles. On a public repository only the check is posted, and details stay in the dashboard.
Agent pushes are billed reviews
Reviews start with each push, not with pull requests. An agent that pushes after each small fix starts a review each time, and after November 10, 2026 each one is billed.
Closed is not the same as fixed
Vercel closes a finding when a later review no longer detects it, and says that does not confirm the issue is fixed. Check the code or the commit that closed it.
It reads code, not the running app or your database
Security Review analyzes source code. It does not test the deployed app, and it cannot see settings that live outside the repository, such as Supabase row level security changed in the Supabase dashboard.
Related guides
Official docs behind this checklist
Browse all launch guides · Read resources · How to make and check your next change
FAQ for builders before launch
Which plans get Vercel Security Review?
Pro and Enterprise, in public beta since November 3, 2026. A team Owner turns it on in the Agent settings.
Is Vercel Security Review free?
Until November 10, 2026. After that each review is billed at the Vercel Agent rate: provider tokens at cost plus $0.25 per million tokens.
Does it replace a security audit?
No. Vercel says it is not a security audit, does no runtime or penetration testing and can miss issues.
How does a coding agent read the findings?
With Vercel CLI 65.10.0 or later: vercel security vulnerabilities --json lists them, and inspect shows one in full. A finding can also be copied as Markdown or JSON from the dashboard.