Vercel Agent

Vercel Agent is the AI agent inside the Vercel platform. It answers questions about your projects, investigates production issues, reviews pull requests and carries out changes you approve, using context from your deployments, logs, metrics, configuration, usage and connected repositories. Vercel describes it as the intelligence layer for shipping on Vercel. It opened as a limited beta in June 2025, Code Review entered public beta in September 2025 and Investigations in October 2025. Where v0 by Vercel, the other Vercel AI product, builds apps from prompts, Vercel Agent works on apps that are already deployed.

Code Review runs automatically when a pull request is opened or new commits are pushed, and on demand when you mention @vercel in a pull request comment. It reads the whole codebase for context, generates patches and runs them in a Vercel Sandbox against your real builds, tests and linters, so only suggestions that pass appear in the pull request, where they can be applied with one click. Reviews follow repository guideline files such as AGENTS.md, CLAUDE.md, .cursorrules and .github/copilot-instructions.md. It is tuned for Next.js, React, Nuxt and Svelte but not limited to them, and supports TypeScript, Python, Go and other languages. Investigations analyze Medium and High anomaly alerts, failed deployments, runtime errors and cost spikes, query the relevant logs and metrics, and summarize the likely cause, and an investigation can also be started by hand from the alert. Chat, in the dashboard or by mentioning @Vercel in Slack, answers project questions and drafts plans for changes such as a rollback, a configuration update or a tested pull request. Installation adds Web Analytics or Speed Insights to a GitHub-connected project and opens a pull request.

The agent is read-only by default. Reading sensitive data such as environment variables, changing Vercel resources or writing to GitHub requires approving a scoped plan, and each approval covers only the listed targets and operations. Pull requests come from the Vercel Agent GitHub App on branches that start with vercel-agent/, with the requesting user added as a co-author, and every write is recorded in the team activity history. In Slack it reads the thread for context and can open a Slack code channel for a task. Through Agent Tools it can query installed Marketplace integrations over MCP, currently Neon, Prisma, Supabase, Dash0, Stripe and Mux, and it can install private npm packages using team-shared registry credentials.

Pricing is usage-based with no seat license. Paid work is billed at the Vercel Token Rate of $0.25 per million tokens, covering input, output and cached tokens, plus the model provider inference cost passed through with no markup. This replaced a $0.30 per-request fee on June 30, 2026. Chat includes 50 free messages per team member per day during the beta, shared with Slack, and approving a plan the agent proposes is charged even within that allowance. Observability Plus includes 10 investigations per billing cycle, and Installation is free. Vercel Pro is $20 per month per developer seat with unlimited viewer seats, and Enterprise pricing is custom. Prices are in USD.

The free Hobby plan includes Installation only. Code Review has no included usage, so each review is billed by the tokens it uses, and the Agent dashboard shows the spend, files read and review time for every review. Investigations require the Observability Plus add-on, billed at $1.20 per million events on Pro. On data, Code Review never trains on customer code when a team has data sharing turned off or is on Enterprise. The Vercel Terms of Service, updated October 19, 2026, leave Model Training off by default on paid Pro, where a team can opt in, and on by default for Hobby and trial Pro, and Enterprise is excluded. Vercel holds a SOC 2 Type 2 attestation and ISO 27001 certification and lists GDPR, PCI DSS, the EU-U.S. Data Privacy Framework and TISAX AL2. A HIPAA BAA is a $350 per month add-on on Pro and is available on Enterprise.

Vercel Agent suits teams that deploy on Vercel and want code review, incident investigation and approved operational changes in the platform that runs their apps and in Slack. Teams that want a pull request reviewer independent of their hosting platform, across GitHub, GitLab, Bitbucket and Azure DevOps, should consider CodeRabbit or Qodo. Teams that want an agent writing code in their own editor or terminal should consider Cursor or Claude Code.

Current state Q4 2026: Chat, Slack, Code Review and Investigations are in public beta on Pro and Enterprise, and Installation is in public beta on all plans. The agent expanded from code review and alert investigations to dashboard chat and approved actions on June 30, 2026, reached Slack on August 19, 2026 and Slack code channels on August 20, and gained private npm package installs on September 30. Pricing is the $0.25 per million token Vercel Token Rate plus provider inference.