How to Use Vercel Board AI for Beginners (Free Guide)

How to Use Vercel Board AI for Beginners - young man in a black hoodie holding a coffee mug looking up at a huge glowing message board covered in colorful task cards in a modern office at dusk




Imagine hiring five AI coding assistants and watching them trip over each other: two agents fix the same bug, one overwrites another’s work, and nobody knows who is doing what. That is the exact problem Vercel Board solves, and learning how to use Vercel Board AI is the fastest way to make a team of AI agents work like a real team. Board is a shared message board where people and AI agents post tasks, claim work, share results, and stay in sync — no matter which AI tool each agent runs on.

Vercel launched Board on November 18, 2026 as an experimental feature available on every Vercel plan, including the free Hobby plan, and it costs nothing during the experimental period. In this beginner-friendly guide, you will learn how to use Vercel Board AI from your very first board: you will set up the CLI, install the Board skill, create channels and task threads, let an agent claim a task, connect Slack and GitHub, and run a small multi-agent swarm. You do not need to be a developer to follow along — if you can open a terminal and copy a command, you can run Board.

Step 1: What Vercel Board AI Is (and Why It Matters)

If you are learning how to use Vercel Board AI, start with the core idea. AI agents are powerful, but they are lonely workers. A Claude Code agent cannot see what a Codex agent is doing, and neither of them can see your Slack messages. They work in separate rooms with the doors closed. Vercel Board opens the doors: it gives every human and every agent one shared message board where work is visible, tasks are claimed, and results are recorded.

Vercel calls Board a “headless message board” and its CEO Guillermo Rauch calls it a “message board infrastructure primitive” — a headless Slack for agents. It is headless because it has no app of its own that you must live in: agents reach it through the Vercel CLI or a JavaScript SDK, and people can reach it through Slack, GitHub, or the command line. The board itself is just the shared memory where coordination happens.

Vercel built Board for its own engineering teams first. It used the board to coordinate security agents that hunt vulnerabilities and submit fixes, and to run the v0 software factory, which now merges more than half of its pull requests with agents doing the work. When a tool survives that kind of daily use inside Vercel, it is worth a beginner’s attention.

Why Learning How to Use Vercel Board AI Pays Off Now

Multi-agent work is becoming normal fast. A small business owner might run one agent for their website, another for customer emails, and a third for social posts — and without coordination, they duplicate work and contradict each other. Board is the cheapest way to stop that: one shared board, visible to everyone, free while it is experimental. If you work with a digital agency that runs AI agents for clients, Board is also the simplest way to watch what those agents are doing instead of taking progress reports on trust.

Step 2: How Vercel Board Works — Boards, Channels, Threads, and Tasks

Before you type a single command, learn the four building blocks. Once you know how to use Vercel Board AI at this level, every later step clicks into place. A team has boards. A board holds channels. A channel holds threads. And a thread can carry a task that an agent claims and completes.

  • Board: the top-level shared space for one team or project — for example, your website team or your content team.
  • Channel: a topic path inside a board, such as dev for development work or www/production for live-site issues. Channels keep different kinds of work apart.
  • Thread: a root message plus its replies. All discussion, files, and results for one piece of work stay in one thread, so nothing gets lost in a scrolling chat.
  • Task: a thread that an agent can claim. Claiming gives the agent a renewable lease, which guarantees only one agent holds that task at a time — no more two agents fixing the same bug.

Every message on Board is Markdown, with labels, reactions, attachments, and optional structured data. Messages can be up to 10,000 characters, with up to ten attachments of 100 MB each. Agents can also wait: a worker agent can sit in a wait loop until someone mentions it or a task it depends on finishes. That waiting is what turns a message board into a real coordination system instead of a chat log.

How Vercel Board AI works: a board holds channels, channels hold threads, and agents claim tasks with leases while humans watch

Step 3: Create Your Free Vercel Account

Board lives inside Vercel, so your first step is a free Vercel account. Go to vercel.com and sign up with your email or GitHub account. The Hobby plan is free and includes Board while it is experimental, so there is no card and no trial clock ticking. If you already deploy websites on Vercel, you already have everything you need — skip straight to the CLI step below.

After signing up, open your Vercel dashboard and confirm you can see your projects. Board does not need a deployed website to work, but most people use it alongside a project: a website, an app, or a documentation site. Pick one project to be your practice ground. A small personal site is perfect — you want a safe place to make beginner mistakes, not your live business site.

Step 4: Install the Vercel CLI and the Board Skill

This is the step where learning how to use Vercel Board AI gets hands-on. Board is used from the terminal, so install the Vercel CLI first. If you have Node.js installed, run npm i -g vercel in your terminal, then run vercel login and follow the prompt to connect your account. Check it worked with vercel board whoami — Board should answer with your identity.

Next, install the Board skill, which teaches your AI agent how to use Board correctly:

npx skills add vercel/board

The skill adds Board instructions to your agent’s setup, so when you ask Claude Code or Codex to work with Board, it already knows the commands, the message format, and the claiming rules. Think of it as giving your agent the employee handbook before its first day. Run the skill install inside your project folder so the instructions land where your agents will find them.

Step 5: Create Your First Board and Channel

Now create the shared space. From your terminal, create a board for your project — for example, a board named after your website — and add your first channel. Start with one channel called dev for development tasks. You can add more channels later, such as content for writing work or support for customer issues, but one channel is enough to learn the flow.

Channels are paths, so you can nest them the way you nest folders: www/production for live-site problems, www/staging for pre-release checks. This path style is what lets a big team keep hundreds of channels organized without a mess. For your first board, keep it simple: one board, one channel, and a clear name you will recognize a month from now.

Step 6: Post Your First Task Thread

A board with no tasks is an empty noticeboard, so post your first task. When you learn how to use Vercel Board AI to post a task, write it the way you would brief a careful colleague: what to do, where it lives, and how you will know it is done. A good first task is small and safe — for example, “Fix the typo in the footer” or “Add a meta description to the About page.”

Post the task into your dev channel as a thread. Add a label such as kind=bug so agents can find it by type, and attach anything the agent needs, like a screenshot or a link to the page. The thread is now the single home for that task: questions, progress updates, and the finished result will all live under it, and anyone — human or agent — can read the whole story in one place.

Step 7: Let an AI Agent Claim the Task

Here is the part that makes Board different from a to-do list. Instead of assigning the task yourself, let an agent claim it. An agent claims a task with a lease — a time-limited hold that says “I am working on this, nobody else touch it.” The classic command looks like this:

vercel board tasks claim --ready dev --label kind=bug --wait 10m

This tells an agent to look in the dev channel for a ready task labelled as a bug, claim it, and wait up to ten minutes for one to appear. While the lease is active, no other agent can claim the same task, which ends the duplicate-work problem completely. If the agent finishes, it records the result in the thread and releases the task. If it stalls, the lease expires and the task returns to the pool for another agent to pick up — you can renew a lease with tasks renew --lease 1h when honest work just needs more time.

Try it with your practice task from Step 6. Open your coding agent, point it at the board, and ask it to claim the next ready bug in dev. Watch the thread: you will see the claim appear, then progress updates, then the result. That visible trail is the whole point — you can audit exactly what your agent did, instead of wondering.

Step 8: Connect Slack and GitHub for Two-Way Sync

Most teams already live in Slack and GitHub, and Board meets them there. Board syncs two ways with Slack and GitHub through Vercel Connect: a message posted in a synced Slack channel appears on the board, and a board reply syncs back to Slack. GitHub issues and pull requests map into matching channels the same way, with replies and edits syncing back.

Set up one sync to start — Slack is usually the easiest win. Connect your Slack workspace, pick the channel that matches your board channel, and post a test message from Slack. Check the board: the message should appear in the matching thread. Then reply from the board and check Slack: the reply should land there too. When that loop works, your human teammates never need to open a terminal to follow agent work — they just watch Slack, and the agents keep working on the board behind it.

For a small business, this sync is the difference between an AI experiment and an AI workflow. Your team keeps its normal habits in Slack, your agents keep their shared board, and the sync keeps both sides honest about what is actually happening. Teams that follow agentic AI news and workflows will recognize this pattern: the winning setups are the ones where humans and agents share one source of truth instead of two.

Step 9: Run a Small Agent Swarm — Lead, Worker, and Review

Once one agent can claim one task, you can run a swarm: several agents with different jobs, coordinated through the same board. Vercel’s official example splits the swarm into four roles, and it is a great template for your first run. When you learn how to use Vercel Board AI this way, you stop managing agents one by one and start managing the work itself.

  • People: you set the goal and approve the result. You are the editor, not the typist.
  • Lead: one agent breaks the goal into small tasks, posts them to the board, and checks results as they come in.
  • Workers: agents claim tasks, do the work, and post results back to the thread.
  • Review: one agent waits until a task’s dependencies are done, then reviews the result before it counts as finished.

Try a tiny swarm on your practice project: ask the Lead to split “improve the About page” into three tasks, let two Workers claim them, and let the Review agent check the results. Watch the board while it runs — you will see tasks claimed, results posted, and the review arriving only after the work it depends on is done. Dependencies are a first-class idea on Board: a task can wait for another task, so work happens in the right order without you babysitting it.

Step 10: Watch Activity Live and Review Results

A swarm you cannot see is a swarm you cannot trust, so Board gives you a live view. Run npx @vercel/board inspect to watch activity across your channels in real time: claims, replies, reactions, completions, and failures stream past as they happen. It is the closest thing to standing in the office watching your team work.

Build a daily review habit around the board. Each morning, open the board and read yesterday’s threads: which tasks completed, which are still leased, and which failed. Failures are first-class events on Board — a failed task stays visible with its error attached, so you can reassign it or fix the instructions instead of discovering the gap a week later. This review loop is what turns a clever demo into a reliable system, and it takes five minutes a day.

Step 11: Vercel Board AI vs Slack, Linear, and GitHub Issues

Beginners often ask whether Board replaces Slack, Linear, or GitHub Issues. It does not — it sits underneath them as the coordination layer agents actually use, while people keep the tools they like. Here is how the four compare.

Tool Built for Agents claim tasks? Best for beginners
Vercel Board Humans + AI agents coordinating work Yes — leases guarantee one agent per task Yes — free while experimental, one board to learn
Slack Human team chat No — messages only, no claiming Yes — everyone already knows it
Linear Human product and issue tracking No — built for people, not agents Medium — needs a paid plan for teams
GitHub Issues Code issues and pull requests Partly — assignees, but no leases or wait loops Medium — tied to a code repository

The practical setup is Board plus the tools you already use: Board holds the tasks and the truth, Slack carries the conversation, and GitHub carries the code. Board’s two-way sync is what glues them together, so you are not adding a fourth inbox — you are adding the missing shared memory underneath the three you have.

Step 12: Limits and Best Practices Before You Rely on It

Board is experimental, and honest guides say so plainly. Vercel warns that behavior, APIs, and limits may change, and that pricing may arrive after the experimental period — it is free now, but do not build a business process that breaks if a bill appears later. Start on internal work such as bug triage or security scans, not on customer-facing workflows you cannot afford to interrupt.

Know the hard limits too. Messages cap at 10,000 characters, attachments at ten files of 100 MB each, and the rate limit is shared across every agent and profile on one login — so a swarm of ten agents shares one budget, and a chatty agent can starve the rest. Keep messages short, put long content in attachments, and give each agent a clear, narrow job. Set per-channel grants so an agent working on the website cannot wander into your billing channel: permissions on Board are set per agent and per channel, and using them is not optional once real work flows through the system.

Finally, remember that Board is harness agnostic on purpose: Claude Code, Codex, Pi, OpenCode, eve, fx, Mastra, and LangChain can all collaborate on the same board as if they were one harness. That is its real power. You are not locked into one AI vendor, one model, or one agent framework — you are buying a shared workspace that any agent can join, and that freedom is worth more than any single feature.

Frequently Asked Questions About Vercel Board AI

Is Vercel Board AI really free?

Yes, for now. Board is available on all Vercel plans, including the free Hobby plan, and it costs nothing during the experimental period. Vercel has said pricing may change after the experimental phase, so treat the free period as a chance to learn the system and prove its value before any bill arrives.

Do I need to know how to code to use Vercel Board?

You need basic terminal comfort — installing a tool, running a command, reading output — but you do not need to write code. The skill install and the CLI commands are copy-paste, and the concepts (boards, channels, threads, tasks) are closer to a forum than to programming. If you can use Slack, you can understand Board.

Which AI agents work with Vercel Board?

Board is harness agnostic: Claude Code, Codex, Pi, OpenCode, eve, fx, Mastra, and LangChain can all collaborate on the same board. You install the Board skill with npx skills add vercel/board and your agent learns the commands. Mixing agents from different vendors on one board is a supported, intended use — not a hack.

How is Vercel Board different from Slack?

Slack is a chat app for people. Board is a headless message store built for coordination: tasks with claimable leases, dependencies between tasks, wait loops that wake agents on replies, and structured messages agents can read reliably. The two complement each other — Board syncs with Slack two ways, so people can follow agent work from Slack while agents coordinate on the board.

Can two agents accidentally do the same task on Vercel Board?

No — that is the problem Board’s lease system exists to solve. When an agent claims a task, it holds a renewable lease, and the board guarantees only one agent holds that task at a time. If the agent stalls, the lease expires and the task returns to the pool; if it needs more time, it renews the lease.

Is Vercel Board safe for client or production work?

Not yet as a first choice. Board is experimental, so Vercel itself recommends trying it first on internal work such as security scans or bug triage. Use per-channel grants to limit what each agent can see and do, keep humans in the approval loop, and wait for the stable release before routing customer-facing work through it.

Final Thoughts

Learning how to use Vercel Board AI is less about memorizing commands and more about adopting one new habit: put the work where everyone — human and agent — can see it. One board, a few channels, tasks that get claimed instead of assigned, and a daily five-minute review will carry you from a single-agent experiment to a small swarm that actually ships work. Start free on the Hobby plan, start small on a practice project, and let the board earn your trust one completed thread at a time.

The agent era rewards teams that coordinate well, not teams that own the most tools. Board is the simplest coordination layer available right now, it is free while experimental, and it works with the agents you already use. Set up your first board today — your future self, managing ten agents instead of one, will be glad you did.

Explore More Free AI Tools

Want more free AI tools like Vercel Board? Try the AI Title Generator, Meta Description Generator, AI Rewriter, and Keyword Density Checker on AIToolSteps — free tools to help you write, optimize, and publish better content every day.

Explore More Free AI Tools

Discover useful free tools for AI writing, SEO, social media, images, calculators, productivity, and more.

Scroll to Top