Vercel Security Review: What a Per-Push Scan Covers and Costs

Push a change to a Vercel-connected GitHub repository and, within seconds, an agent has read the diff and decided whether you just shipped a missing authorization check. That's what Vercel promises with Security Review, a new part of Vercel Agent that Vercel announced Nov. 2 and put into public beta for Pro and Enterprise teams on Nov. 3, 2026.

The headline is short: scan on every push, fix in one click. The launch post, the changelog entry and the Vercel Agent docs say more, including a price and a date that matter if you're about to switch it on.

Vercel's launch post on X, Nov. 3, 2026 (@vercel)

View on X

The post reads: "Today we are launching Vercel Security Review in public beta. Run vulnerability reviews on every push, and ship fixes in one click with Vercel Agent."

It reads the diff, not the repo

Security Review starts from GitHub push events rather than pull requests, so direct pushes and force pushes to your production branch get reviewed too, according to the changelog. When one push touches several projects, they're reviewed together. In monorepos, the launch post says, a review is scoped to the project's root directory.

Each review "compares the push against a base and covers only what changed." For deployments, the base is the branch's previous deployment. The post says that's why it "finishes in seconds and can run on every preview and production deployment."

What the agent does with that diff is the interesting part. Per the post, it reads the changed code with the code around it, traces how the change can be reached, checks for existing protections and rechecks each finding before reporting it. Vercel's examples are logic flaws such as missing authorization checks, plus exposed secrets, and it says pattern-matching scanners miss the first kind.

Our reading: a diff-only scan is cheap and fast, but it can only flag what a push introduces. A flaw already sitting in your main branch before you enabled the feature isn't part of any later diff. Neither the post nor the docs describe a full-repository scan, so treat this as a guard on new changes, not an audit of old code.

Seven days of free

The pricing paragraph is the one most summaries skip. Security Review is billed "at the Vercel Agent rate": provider model tokens at the provider's rate with no markup, plus $0.25 per million tokens, according to the post and the changelog. The changelog adds that cost "scales with how often you push and how much code changes."

Reviews are free until Nov. 10, 2026. That's a week from the announcement. Fixes made with Vercel Agent are billed at the Agent rate from the start.

Vercel doesn't publish how many tokens a typical review uses, so we can't estimate a monthly bill, and we won't invent one. The way to get a number is to enable it on one busy project during the free window and read the usage afterward. Two inputs set the cost: how many pushes your team makes, and how large each diff is.

Where a finding goes

The changelog says each finding records the project, environment, branch and commit, with a summary, a suggested fix and a severity rating that the system assigns automatically, not a person. Findings land in the Vulnerabilities dashboard. On private and internal repositories, Security Review also adds a GitHub check to the commit and comments on the pull request. On public repositories, "finding details stay in the dashboard."

Every new finding is checked against earlier ones so the same issue isn't flagged twice, and when a later review no longer sees it, the dashboard marks it "no longer detected." Teams can write security policies in plain language for all projects or specific ones. The post's example: any endpoint that returns customer data must check that the requester owns it.

For tooling, vercel security vulnerabilities pulls findings into your own coding agent and needs Vercel CLI 65.10.0 or later. Findings can also be copied as Markdown or JSON. Each new finding creates an audit log event, vercel-agent-security-finding-created in the audit log docs, and Enterprise teams can send those events to a SIEM through Audit Log Drains.

One click, then a pull request

On a finding, "Fix with Vercel Agent" hands it to the agent. The post says the agent first confirms the vulnerability still exists, then makes "the smallest change that fixes it" and updates the tests. The changelog is explicit about what happens next: the agent opens a pull request for you to review, and "nothing merges until you approve it."

So "ship fixes in one click" means one click to a proposed pull request. The merge is still yours. The post adds that suggested fixes you can apply directly from the pull request are something Vercel is working on, which means that isn't available now.

What the docs say it isn't

The limitations paragraph is short: "Security Review analyzes source code. It isn't a security audit, and it doesn't perform runtime or penetration testing." It "helps you find vulnerabilities but doesn't guarantee your application is free of them."

Three other limits show up when you read the pieces together:

• GitHub only, for now. The post lists repositories that don't deploy on Vercel, and deployments from the Vercel CLI and v0, as next steps. That suggests they aren't covered today.
• Plans. Pro and Enterprise teams. The X post doesn't say so; the docs do.
• Beta. Vercel hasn't published detection rates, false-positive rates or which models run the review. Unconfirmed.

Four quick answers

Does it scan my whole repository? Not according to Vercel: each review covers the push against a base. Unconfirmed whether any full scan exists.

Is it free? Until Nov. 10, 2026. After that, tokens plus $0.25 per million.

Will the agent merge fixes on its own? No. It opens a pull request, and the changelog says nothing merges until you approve.

Does a clean review mean my app is secure? The docs say it doesn't guarantee that.

Before Nov. 10

If you turn it on this week, pick one project with frequent pushes, not your quietest repository. Compare what Security Review reports against whatever scanner you already run, and note which findings only one of them caught. When the free window closes, you'll know what a week of real pushes cost, and your own bill, not the launch post, is what decides whether it stays on.

Source: Vercel, Introducing Vercel Security Review

vercel vercel security review vercel agent code security github

AI

About the author

AI Briefing Daily Editorial. The editorial desk of AI Briefing Daily, an independent publication covering artificial intelligence. Every story is researched against primary sources, checked by hand and reviewed by a person before publication.