Vercel Agent

Vercel Agent is Vercel's built-in AI agent for reviewing code and operating deployed applications. Where v0, Vercel's other AI product, builds apps from prompts, Vercel Agent works at the other end of the lifecycle: it reviews pull requests, investigates production issues, answers questions about your projects, and takes approved actions, all from inside the Vercel dashboard. It is positioned as the intelligence layer for shipping on Vercel. Its two core capabilities are code review and production investigation. For code review, the agent analyzes pull request diffs in the context of your whole application to catch security and performance issues and recommend fixes. Critically, it does not just suggest: each proposed patch is run and validated in an isolated Vercel Sandbox against your real builds, tests, and linters, so only high-signal, ready-to-ship suggestions surface in the pull request rather than the noisy output common to review bots. For production investigation, you can point the agent at a failed deploy, a runtime error, or a cost spike and it traces the root cause, separates genuine incidents from normal activity to cut alert noise, and provides remediation steps tailored to your stack. Vercel Agent is framework-agnostic. It is tuned for popular frameworks including Next.js, React, Nuxt, and Svelte, but is not limited to those, a meaningful contrast with v0's focus on the Next.js and Vercel stack. It connects to your Git repository for pull request workflows and draws on Vercel platform observability for production data. Pricing is usage-based rather than per seat. There is no seat license: Vercel Agent is available to every member of a team on the Pro or Enterprise plan, billed at the Vercel Token Rate plus underlying provider inference costs, which pass through at cost. Vercel plans are Hobby at $0, Pro at $20 per user per month, and Enterprise at custom pricing. Because cost is consumption-based, spend scales with how much review and investigation work the agent performs, which is transparent but harder to forecast than a flat per-seat fee, and there is no single published all-in per-unit rate for the agent itself. On data handling, Vercel Agent does not train on customer code when a team's data-sharing setting is turned off, and never trains on code for teams on the Enterprise plan. It operates under Vercel's platform compliance, which includes SOC 2 Type II and GDPR. The agent is read-only by default; any remediation or code change it proposes requires human approval before it is applied. Vercel Agent is not the right fit for teams that do not host on Vercel, since its deepest signal comes from Vercel-hosted production data and its workflow is built into the Vercel dashboard. Teams wanting a standalone code-review tool decoupled from a hosting platform, or a general coding assistant, should evaluate Cursor ($20/mo) or Claude Code ($17/mo billed annually); teams focused on building applications from prompts should look at v0. Current state Q3 2026: Vercel Agent is generally available on Pro and Enterprise plans with no seat license, having expanded in July 2026 from code review into a broader dashboard agent that also investigates production issues, answers project questions, and takes approved actions. Runs and patches are validated in isolated Sandboxes. Pricing is the usage-based Vercel Token Rate plus provider inference pass-through.