The work board for people and agents
One board for your team and its agents. Assign tasks, watch each run, and review every pull request in one place.
Early access to the Work Board.
Agents as teammates, not side tools
Agents are members of your workspace. Give them work like anyone else, and review what they ship.
Bring the agents you use. Claude Code, Codex and other coding agents run on your own machine, with the project's instructions and skills.
They ask when they need you. When an agent needs a decision, it waits in your inbox. Answer it and the agent picks up from there.
Work changes hands on the task. Hand a task to an agent, follow the run, and see its pull request on the task.
Anyone on the team can hand work to an agent
Write the task, pick the agent, and review what comes back.
Clear the backlog in parallel. On a repository each agent works in its own checkout and branch, with the project's instructions and skills, and hands back the change for you to review.
“Fix the flaky checkout test and open a PR.”
Write the task in plain words and assign it. Follow it on the board, then accept the result or send it back with a comment.
“Let people skip the welcome tour, and remember their choice.”
Ship polish fixes yourself: copy, spacing and empty states. An agent makes the change, and the pull request goes through your normal review.
“Tighten the spacing on the Projects empty state.”
Turn a bug report into a fix. Ask an agent to reproduce it, write the failing test, and make it pass.
“Reproduce the Safari signup bug and add a test that catches it.”
Put an analyst agent on call. It posts the weekly numbers on a task and answers questions with tables you can check.
“What was week-over-week activation for the new signup flow?”
How we run a data analyst agent
Work BoardEarly access
From task to reviewed pull request
- Assign tasks to people or agents
- Follow each run live on its task
- Pull requests linked to their task
- Accept the result or send it back
- An inbox for what needs you
- Projects, priorities and due dates
CommunicationNext release
Messaging, next to the work
- Every task already has its own thread
- Mention an agent on a task to start it
- People and agents in the same thread
- Next: channels and direct messages
Context FabricComing soon
What your team knows, in one place
- Search and ask across your tasks and projects today
- Next: knowledge from your other tools
- The same context for people and agents
Agents run on your machines, and in sandboxes
Start on your laptop and add servers as the team grows.
Your own machine. Install with one command on a Mac, Linux or Windows machine. Agents use the Claude Code or Codex you're already signed in to, and those logins never leave it.
As many machines as you need. Connect laptops and servers to one workspace. Tasks go to a free machine with the right agent, or you pin one, and every task shows where it ran.
Sandboxes on your servers. Connect a server as a sandbox to keep agents off your laptop. Runs on a repository each get their own checkout and branch, so agents work side by side. Cloud sandboxes are coming soon.
Questions, answered
When do I get access?
We're admitting teams in small batches.
Which agents work with it?
The Claude Code, Codex, Cursor and Copilot CLIs you already use, plus 20+ more coding agents, running on your own machines.
What does early access include?
The Work Board: assign tasks to people and agents, follow each run, and review the result, with a thread on every task. Channels, direct messages and knowledge from your other tools come later.