An AI coding agent can now take a Vercel domain purchase from name search through the final buy command. The important boundary is where that command runs: in a non-interactive session, vercel domains buy returns a structured error and suggested next commands instead of completing the purchase. A person must review and confirm the transaction in an interactive terminal or the Vercel dashboard.
That changes the useful delegation question. You can ask an agent to research names and prepare the purchase, but what exactly should it bring back, and what should you verify before you authorize it? Vercel’s October 9, 2026 changelog describes the new sequence. Its CLI documentation lists the relevant commands and the separate steps for attaching and verifying a domain.
The flow has four commands, each with a distinct job:
vercel domains search <query> finds candidate names from a keyword or domain fragment.vercel domains check <domain> checks whether a candidate is available.vercel domains price <domain> returns pricing for one or more domain names.vercel domains buy <domain> initiates the purchase workflow.
This lets an agent do the repetitive discovery work: explore naming variations, check availability, and assemble the pricing information for your shortlist. Vercel says the CLI also ships an agent-facing skill for the sequence, so the agent can be guided through discovery, pricing, and purchase without you spelling out each command.
Treat the first three commands as research. Ask for the exact domain string, its availability result, and the returned purchase and renewal information. If the domain is unavailable or the pricing output is unclear, have the agent stop and report that; don’t let it silently substitute a different name. The choice of a public name is a product decision, even when the agent can search faster than you can.
Vercel’s safeguard is tied to how the purchase command is run. In a non-interactive environment, vercel domains buy does not finish the purchase. It returns a structured, machine-readable error and suggested next commands, handing the confirmation back to the user. Vercel describes this as preventing the agent from spending money without approval.
You can then complete the confirmation in an interactive terminal or in the dashboard. Before you do, check three things Vercel calls out directly: the price, whether automatic renewal is on, and the registrant contact details. Read the values shown at confirmation time; don’t rely on a number or contact record copied from an earlier search.
That boundary is useful, but it does not make the review optional. The agent can still find a name you did not intend, select a confusingly similar spelling, or report a candidate that is unsuitable for your product. Human confirmation is the point where you verify that the exact string, account scope, renewal choice, and registration details match your intent.
Ask the agent to return a compact handoff before it attempts the final command:
Then compare that handoff with the confirmation screen. If the domain, price, renewal setting, contact details, or account scope differs, stop and resolve the mismatch before confirming. A domain registration is a real purchase attached to an account; a prompt saying “buy something suitable” is not a substitute for checking the final transaction.
You can also set the boundary in your task instructions: have the agent search and compare candidates, but require it to stop before completing the interactive confirmation. The non-interactive behavior is the technical backstop described by Vercel; clear task wording helps keep the intended handoff visible in the conversation.
The domains CLI reference describes vercel domains buy <domain> as buying a domain for the current scope. It separately documents vercel domains add <domain> <project> to add a domain to the current scope or a Vercel project, and vercel domains verify <domain> to check DNS configuration. In other words, the purchase command and the project-connection workflow are distinct operations in the CLI.
After the registration is confirmed, decide whether and where the domain should be connected. Check the selected project, then use the documented add and verification steps for that project. Do not let a successful purchase stand in for a verified connection: the registration can be complete while the app’s domain configuration still needs attention.
This distinction also helps you keep the agent’s authority narrow. It can prepare a purchase handoff without being asked to change project configuration, DNS, or deployment settings. Add those tasks only when you want them, and review the expected domain and project before they run.
For a new app, use a staged request: This puts the agent on the preparation side and you on the purchase decision. If you do not want to buy anything yet, stop after search, availability, and pricing; those commands are separate from the purchase command in Vercel’s CLI reference.
The update is not a domain strategy or a deployment shortcut. It is a command-line workflow that lets an agent prepare a purchase and carry it to a deliberate human confirmation. That can remove context switching during setup, while keeping the paid commitment visible at the moment it happens.
If you are already working through how to move from a preview URL to a domain you own, see the custom-domain guide. For a separate example of agent-assisted payments in a different product workflow, see the Replit payment walkthrough. OTF kits provide application code; they do not provide a Vercel domain-purchase command or a deploy command. Use Vercel’s CLI documentation for this workflow, and verify the commands available in your own project before asking an agent to do operational work.