Skip to content
Get packs

Guide · Agent skill

How to use the Vercel CLI agent skill (vercel-cli)

Updated · written with AI from the project's own documentation and checked against our catalogue data

The vercel-cli skill teaches an AI agent to publish web applications to Vercel using the vercel command-line tool, and to manage projects, domains and environment variables. It is written for people who already have a Vercel account and want an agent to push finished projects live and check the result. After this guide you will know what the skill does, how to install it, how to ask for a deploy in plain language, and what to check before trusting it.

Vercel Deployment CLInanocoai/nanoclaw · MIT

Quick start

  1. 1Download and unzip the skill. Download the ZIP from this page and unzip it so the folder contains SKILL.md.
  2. 2Place the folder or upload it. Put it in ~/.claude/skills/ for all projects or in .claude/skills/ for one project, or upload the ZIP under Settings → Capabilities → Skills in the Claude apps.
  3. 3Install the Vercel CLI and Node.js. Make sure the vercel command, Node.js and npm or pnpm, and curl are available on the machine where the agent runs.
  4. 4Register your Vercel token. Create a token in your Vercel account and register it with OneCLI so the proxy can inject it into requests.
  5. 5Ask for a deploy. Describe the project and whether it should go to production or be a preview in plain language.
  6. 6Review the result. Check that the agent reports a READY URL that returns a 2xx or 3xx response before you rely on it.

What is an agent skill and what does vercel-cli do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a request matches the skill's description. This one is called vercel-cli and its description says it should be used when you ask to deploy, ship or publish a web application, or manage Vercel projects, domains and environment variables.

Once active, the agent runs a fixed checklist: it verifies that its Vercel login works, builds the project locally, deploys it, and confirms the live link responds before sharing the URL. It can also list recent deployments and projects, add domains, and pull or add environment variables.

Install the vercel-cli skill

Download the ZIP from this page and unzip it. The skill is a single file, SKILL.md, so you can place the folder in one of two locations. For all your projects, put it in ~/.claude/skills/. For one project only, put it in that project's .claude/skills/ folder.

You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

  • Personal install: ~/.claude/skills/vercel-cli/
  • Project install: .claude/skills/vercel-cli/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Set up the requirements

The skill needs the Vercel CLI (the vercel command), Node.js with npm or pnpm for the local build step, and curl for the live URL check. Every command also needs a Vercel account and access token. In the NanoClaw setup the skill expects the token to be registered in OneCLI, with the OneCLI proxy injecting the real token into requests.

The optional agent-browser skill enables a visual check of the deployed page. The optional frontend-engineer subagent is only needed if you want the agent to hand website builds to a separate agent.

  • Vercel CLI (vercel command)
  • Node.js and npm or pnpm
  • A Vercel account and access token
  • OneCLI with the token registered (NanoClaw setup)
  • curl
  • Optional: agent-browser skill for visual checks
  • Optional: a frontend-engineer subagent for website builds

How to trigger the skill with natural requests

You do not need to type a command. Describe the outcome in plain language and include the project location if it is not the current folder. The skill's description lists deploy, ship, publish, and managing projects, domains and environment variables as triggers.

Be specific about production versus preview. The skill's commands use a production deploy by default for the main example, and a preview deploy without the --prod flag.

  • "Deploy the project in /home/me/landing-page to production."
  • "Ship a preview deployment of this app and send me the link."
  • "List my recent Vercel deployments."
  • "Add example.com to the current Vercel project."
  • "Pull the environment variables from Vercel into a local .env file."

What the agent runs and checks before reporting success

For a deploy, the skill tells the agent to first confirm login with vercel whoami, then run a local build, then deploy with the vercel CLI. It checks that the output shows a Production URL with status READY, using vercel inspect, and that the live URL returns a 2xx or 3xx status with curl. If a check fails, the agent should fix the problem and redeploy before reporting back.

The skill also lists common errors and fixes, such as a missing build script, rate limits, plan project limits and network problems to api.vercel.com.

  • Confirm login: vercel whoami --token placeholder
  • Local build passes before deploy
  • Deploy reports a READY production URL
  • Live URL returns HTTP 2xx or 3xx
vercel whoami --token placeholder vercel deploy --yes --prod --token placeholder vercel inspect <deployment-url> --token placeholder curl -sI <url> | head -1

Safety notes: scripts, network and credentials

This skill runs scripts and uses the network. The agent executes the vercel CLI, npm or pnpm builds and curl on your machine, and sends requests to Vercel. Every command in the skill passes a placeholder token because the real token is supposed to be injected by the OneCLI proxy, so the real secret should never appear in the agent's commands.

Deploys and environment variable changes are real actions on a live service. Review what the agent is about to do, especially for production projects, before letting it run unattended.

  • Keep the Vercel token in OneCLI rather than in project files
  • Review production deploys and env var changes
  • Do not use this skill on sensitive production environments without review

Tips for using the vercel-cli skill well

Run the build yourself or ask the agent to run it first, so build errors show up locally. Use the --cwd option rather than changing directories, which the skill's best practices recommend to keep the working directory stable.

For Next.js projects the skill says vercel deploy detects the framework automatically. Only add a vercel.json when you need custom build settings, rewrites or headers.

  • Build locally before deploying
  • Use --cwd instead of cd
  • Next.js projects are auto-detected
  • Add vercel.json only for custom settings
vercel deploy --yes --prod --token placeholder --cwd /path/to/project

What people use it for

  • Pushing a finished static or Next.js landing page live to a production URL
  • Creating a preview deployment of a branch so a colleague can review it
  • Listing recent deployments to see what is currently live
  • Adding a custom domain to an existing Vercel project
  • Pulling environment variables from Vercel into a local .env file for development

Frequently asked questions

How do I install the vercel-cli agent skill?

Download the ZIP from this page and unzip it. Place the vercel-cli folder in ~/.claude/skills/ for all projects, or in a project's .claude/skills/ folder for one project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

Do I need a Vercel account to use this skill?

Yes. Every command needs a Vercel account and an access token. The skill expects the token to be registered in OneCLI, where the proxy injects the real value into requests.

Is the vercel-cli skill free?

The source is published under the MIT licence. The page does not state pricing for Vercel itself, and the skill notes that project limits depend on your Vercel plan.

Why does the skill use --token placeholder?

The skill says the Vercel CLI needs a token present to skip its local credential check. The placeholder value is replaced with the real token by the OneCLI proxy, so the real secret is not written into commands.

Can the agent build a website for me with this skill?

Not directly. The skill forbids the agent from writing HTML, CSS or JavaScript itself for website builds and instead routes that work to a Frontend Engineer subagent. That subagent is optional and may not exist in your setup, so this skill is best for deploying projects that are already finished.

What should I do if the deploy fails with an auth error?

Run vercel whoami first. If it fails, the skill says to ask for the Vercel token to be refreshed or added in OneCLI, then retry. Also check that the vercel command and network access to api.vercel.com work.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use vercel-cli if you already have a finished web project, a Vercel account, and an agent setup such as NanoClaw with OneCLI, and you want the agent to build locally, deploy and confirm the live URL. Pick something else if you do not have a Vercel token, want the agent to design and code a website itself, or would be deploying to sensitive production without reviewing each action.

Download the skill