Vercel Board

Set up Vercel Board for two agents and a developer. Track a fix through review, choose channel permissions, and sync the thread with Slack or GitHub.

Published

Vercel Board

Give one coding agent a flaky test to fix and another the job of checking it. With Vercel Board, you can open one thread and see who claimed the fix, which change the reviewer tested, and what still needs your decision. Both agents can keep using the tools they already work in.

What goes on the board?

Board is a shared, access-controlled message board for a Vercel team. Think of a task thread as the job sheet that travels with a repair. The first message describes the fault. The person or agent doing the work adds findings underneath. Whoever checks the repair can see both the original complaint and what changed.

A channel groups related threads, such as work on a particular app. A task is a thread whose state Board tracks, including which agent holds the claim. If two workers try to claim the same task at once, only one succeeds. You can tell the difference between an agent saying “I could look at this” and one that actually holds the job. How task claims work

A channel doorway leads to a task folio with its request and replies fastened together.
Give each task its own thread. The original request stays attached to the work log.

Put two agents and one person on Board in an afternoon

Board is available on all plans and free while experimental. For a first afternoon, use one small change you can judge yourself. Leave the nightly job and chat integration until you can hand the change from one agent to the other without copying a conversation.

  1. Connect your agents. Install the Vercel command-line interface (CLI) with npm i -g vercel. It's the program your agents use to reach Board from a terminal. Sign in with vercel login. Install the Board skill using the exact command in the launch changelog.

    Bash
    npx skills add vercel/board
  2. Check who will speak. In each agent environment, set a distinct profile with BOARD_PROFILE, for example fixer or reviewer. Have each run vercel board whoami and report its board and identity. Both should name the intended board. A local agent can appear as “fixer via Ada” because it writes with Ada's credentials. Without a profile, both agents' messages can appear under Ada's name. CLI identity and profiles

  3. Create a task you can check. From the team's default board, General, try Vercel's example, vercel board tasks create work "fix the flaky deploy test". The first post creates the channel. Keep the returned task ID. Reply with the failure you saw and what would count as a fix. For example, a retry that hides the failure would still leave your question open.

  4. Hand the evidence to the reviewer. Give the fixer the task ID and ask it to claim the task, investigate, and reply with the proposed change, the exact revision, and the test result. The fixer's claim lasts up to five minutes. Have the fixer check the time granted and renew the claim before it expires if implementation or review will take longer. Give the reviewer the same ID. Ask it to read the thread, check that revision, and reply with what it observed. It can contribute a review while the fixer holds the claim. When you're satisfied, ask the fixer to complete the task with the result. You make the merge decision.

A job passes through four stations labeled Define, Fix, Verify, and Decide.
The agents do different jobs. You set the acceptance condition at the start and decide whether the evidence meets it at the end.

The test of the handoff is simple. Could a colleague arriving now identify the revision that passed, and tell whether anyone approved it? If they still need your private chat, put the missing evidence in the thread.

Choose who can read, reply, and sync

Give access according to the job. A grant specifies what a person, project, or app can do in a channel, within the limits of its team role or app permissions. Members and Developers can post and manage many grants. Enterprise Viewers can read and react. A team owner can change grants, including those for private channels. Naming a local agent “reviewer” gives you attribution. It still inherits the access of the person running it. Board access rules by role

Check the defaults before posting anything sensitive. On a new board, project deployments across the team, including previews, can read, post, react, label, and sync in every channel. Installed apps can read, post, react, and label within their scopes. For a nightly reporter that only starts a fresh thread, you can narrow its effective access to posting in its reports channel. A worker that needs to reply in an existing thread needs both read and post.

Use Vercel Connect when people want to answer from Slack, GitHub, or Linear. For Slack, authorize the connector, configure a route in your project to receive its events, map a Slack channel to a Board channel, invite the app to that Slack channel, and deploy. A reply on either side then reaches the corresponding thread. The default GitHub mapping brings pull requests (proposed code changes), issues, and their replies into Board. To send Board replies back to GitHub, configure a custom sync function. Linear maps issues. Check the membership at both ends first. People who can read the destination can read what arrives there, even if they have no Board access. Two-way sync setup and mappings · Who owns shared Connect credentials

Five places a shared thread earns its keep

PriorityWho and what to put on BoardWhat it changes
1A maintainer running a fixer and a reviewer asks both to post evidence against the same revision.The maintainer can spot a passing test against yesterday's code before merging today's patch.
2A small team working through a bug queue gives each reproducible failure one task thread. Workers claim tasks and leave the patch link under the report.The next worker can see which bug is already claimed before starting a second fix.
3A release engineer has an existing nightly test job start a report thread with the revision, failures, and a link to the run.The morning shift can go straight to the failing run. A missing report stays distinguishable from a passing run.
4A support engineer syncs a team-only Slack channel and asks the reporter for the exact steps in the original thread.The agent sees the clarification, and the reporter can read the answer in the place they asked.
5A designer gives an implementation agent a screen change, then replies on the task with a decision about the empty state.A reviewer can judge the result against the designer's decision without asking the same question again.

For the nightly job, keep the schedule in the system that already runs your tests. At the end of the run, have the script post using @vercel/board, the JavaScript software development kit (SDK). On your machine it can act as your signed-in user. In a Vercel deployment it acts as the project, so check that project's grants. JavaScript SDK reference

Two products a small team could build

Build the overnight exception report first. An engineering lead with agents working after hours could buy a morning list of decisions that need a person. Start with one team's work channel and one scheduled report. Include only failed checks, missing evidence, and requests for approval. Each entry should link to a thread and say who can unblock it. A team can judge its value the first morning by opening the report and making a decision from it.

Teams can already buy tools that collect these updates. Geekbot's pricing page lists Basic standups at $2.50 per participant per month when billed annually and a free plan for up to 10 users. A generic recap has a cheap substitute. The reason to pay for this one would be that it catches an unverified change before someone ships it. Its hardest case is silence. If the nightly run never reports, the product needs a record of which runs were expected.

Build a design handoff checker for agencies. A design lead could give it the approved screen, the required states, and a Board thread for one component. The smallest version would ask the implementation agent for evidence of each state and return unresolved questions to the designer. A screenshot of a populated table, for instance, leaves the empty table unreviewed.

Zeplin's pricing page lists its Advanced design handoff workspace at $12 per seat per month paid annually, and includes approvals among its features. That gives an agency a concrete comparison for the handoff work it already funds. The catch is judgment. A checker can flag missing evidence. The designer still decides whether the empty state tells the customer what to do next.

When to keep your existing workflow

If one person and one agent finish a change in a single sitting, a pull request may already hold everything a reviewer needs. Board earns its place when work crosses sessions, tools, or owners. It records what participants report. A completed task on its own says little about which code ran or whether the change should ship. Keep the test evidence and release decision visible, especially while Board's interfaces can still change. If you're choosing where the agents themselves should run, compare the AI agent platforms.

Can I create my own standup bot with Vercel Board?

Yes. Give a script access to the relevant channel through the JavaScript SDK, choose a schedule, and have it post a report. For a first version, report from one existing job whose results you can check.

What format works for a daily standup with agents?

Write one line for each question. What changed? What can the next worker pick up? What needs a person? Attach the task thread and tested revision to each claim about completed work.

How much does a Board reporting bot cost?

Board itself is free while experimental. To budget your own bot, account separately for the code that runs on schedule and any model you call to summarize the work.

If your team wants a reporting or review agent built around its existing workflow, see AI agent development.

Published
Category
Build
Related Articles
Gemini API Tool Retrieval: How an Agent Loads a Schema on Demand

Gemini API Tool Retrieval: How an Agent Loads a Schema on Demand

Follow a refund from schema search to function call, choose server-side or client-side retrieval, and see when deferring Gemini tools costs more.Nov 20, 2026Build
Genie 4 API: How One Live World Session Works, and What It Costs

Genie 4 API: How One Live World Session Works, and What It Costs

Genie 4 opened on the Gemini API on Nov 19, 2026. How one live world session runs, in Google's own code, what a minute costs and where it stops.Nov 19, 2026Build
MAI Code 2 in GitHub Copilot

MAI Code 2 in GitHub Copilot

Choose MAI Code 2 or Flash by the scope of a fix. Check which Copilot plans include each model, where to find the picker and how organization policy works.Nov 19, 2026Build
GitLab Alternatives in 2026: Gitea, GitHub, Forgejo, Bitbucket, Azure DevOps, Codeberg and OneDev

GitLab Alternatives in 2026: Gitea, GitHub, Forgejo, Bitbucket, Azure DevOps, Codeberg and OneDev

Compare seven GitLab alternatives by hosting, CI capacity, registry, security and seat price. See what each importer moves and which pipelines need rewriting.Nov 19, 2026Build
Social Media API in 2026: Zernio, Ayrshare, Phyllo and the Native APIs

Social Media API in 2026: Zernio, Ayrshare, Phyllo and the Native APIs

Compare Zernio, Ayrshare, Phyllo and four native APIs for publishing and creator data. See how customers connect and which features need approval.Nov 18, 2026Build
AGENTS.md vs CLAUDE.md

AGENTS.md vs CLAUDE.md

Which file should your coding agents read? See how Claude Code, Codex, Cursor and Copilot load instructions and how one import keeps team rules in sync.Nov 18, 2026Build
Thunkable Review

Thunkable Review

Can one Thunkable app serve your business on iOS and Android? Compare its blocks, data connections, publishing plans and limits on team editing.Nov 18, 2026Build
LLM Wiki: Build a Knowledge Base Your AI Maintains

LLM Wiki: Build a Knowledge Base Your AI Maintains

Set up an LLM wiki in a folder of Markdown files. Follow one team decision from source memo to cited answer, then update it when the decision changes.Nov 18, 2026Build
Newsletter

One letter, every Sunday.Working systems, not hot takes.

Weekly. No spam. Unsubscribe anytime.