Anyone trying Vercel Board for agent coordination?
Vercel Board gives agents shared task threads that sync both ways with Slack and GitHub.
Vercel Board gives agents shared task threads that sync both ways with Slack and GitHub.
One thoughtful comment unlocks the ability to create threads.
I’m pretty excited about this, I think this is roughly how we’re using Linear now with a lot of the old UI that we no longer use stripped out!
In most cases the agent native version of a thing is just the simplest possible UI on top of a robust API (this is probably still better than fully headless because it makes working with humans much easier).
Agreed, and my night-shift dashboard ended up the same way. The agents write to the log through Git, and the dashboard exists for the one human who has to decide what to merge at 7 a.m. The API is where the work happens, and the UI is where I review it.
That’s why Board’s grants interest me as much as the task threads. A thin UI only stays thin if it’s clear what an agent can do without asking a human first.
Have not tried it yet, since we built our own communications tunnels so that various chats and code sessions and so on could all speak to each other in an orderly manner with constantly evolving rules about priorities, work reviews before deploy, hierarchy and assigned tasks.
The idea of agents claiming work, as opposed to having a coordinating agent assign the tasks is interesting. Do you think it's a better use of budget to have the agents claiming the work versus setting up one agent to rule them all, so to speak?
I think having a way for your agents to communicate effectively is important. We created a board for this, where we can bring together agents from Claude, or Codex, or any other system. They work really well together, and they can choose which other agent to talk to and when. It's effective and kinda fun to watch as well.
Vercel shipped Board yesterday as an experimental “agentic primitive.” It’s a shared message board where people and agents post tasks, claim work, and keep each job’s context in one thread, with sync to Slack and GitHub.
I’ve been solving the same problem with less infrastructure. My agents write JSON logs to a private GitHub repo, and a night-shift dashboard reads that repo so I can see what happened overnight when I wake up. Git already handles history and access control. What it doesn’t handle well is letting agents claim work without colliding. Board’s task leases and per-agent permissions look like the pieces I’d otherwise build myself.
Has anyone tried it yet? I’m curious whether it replaces a setup like mine or sits on top of it, and how claiming holds up once you have more than a few agents working at once.