Boards

Boards for the work your agents do.

Every item on a board is a record in your context layer, in a shape your agents read. The column it stands in can start the workflow they follow, and only work that's ready is picked up.

Boards you define

A board is a kanban board for AI agents and the people who direct them: columns you name for one kind of work, with the items standing in them.

Make one per kind of work, with two columns or five named in your own words. Set it up in the web app or ask your agent, and change it whenever your work changes.

Items are your record types

Every item on a board is a record of a type your project defines, and the board's columns are that type's statuses.

Take a bug: its type asks for a severity and the steps to reproduce, and its statuses become a triage board's columns. Bugs and stories can share a board, and you define their types in the library.

Columns start workflows

A column can start a workflow: when an item reaches it, the workflow watching that column starts a run on the item and moves it on when the run ends.

How a workflow is built, from its nodes and loops to every other way to start one, is on Workflows.

Only ready work is picked up

Your agents take an item only when it's ready: written to what its type asks for, blocked by nothing and waiting on no other item.

Among ready items, work for an earlier goal comes first, and work waiting for review can go ahead of new work, so what's half done finishes first. You set how many runs work at once, and each item is taken by one run at a time.

Linked to the roadmap

Any item can serve a milestone, and the roadmap's order decides which ready item your agents take first.

Attach a story, and every task beneath it counts toward the milestone. Goals and what done means live on the roadmap.

Everything waiting on you

When a run on an item needs you, it waits in Actions, the one list of everything waiting on you.

Each item shows the run working on it and how far it has come. Answer in the web app or from your own agent, as Runs & actions shows.

A triage board, from report to fix

A triage board for bugs: a triage workflow files new errors in Reported, and Ready to fix starts an implementation workflow.

  1. An error is reported

    Your error tracker posts a new error to your triage workflow, which files a bug in Reported with what happened and how bad it is. An error it has seen before opens no second item.

  2. Triage writes it up

    The triage run writes down the steps to reproduce and moves the bug to Ready to fix.

  3. The fix starts

    Reaching Ready to fix starts the implementation workflow. Its run takes the bug when it's next in line, and the item shows where the run stands.

  4. Fixing until the checks pass

    On your machine, an agent fixes the bug and tries again until your project's checks pass.

  5. You review it

    The run waits in Actions with the fix, its check results and the branch. You approve it.

  6. The bug moves to Fixed

    The engine lands the change, the workflow moves the bug to Fixed, and its history keeps where it came from, who fixed it and who approved it.

Nobody assigned the bug or chased it: its columns moved it along, and your time went into the one review that needed you.

Questions about boards

Can I set up my own kanban board for AI agents?

Yes. Create a board for any kind of work, name its columns and choose which of your record types stand on it, in the web app or by asking your agent. Change or remove it whenever your work changes.

Can a board start a workflow automatically?

Yes. Set a workflow to watch a column, and every item that reaches the column starts a run. When the run ends, the workflow moves the item to the column you chose for that outcome.

Can bugs and features run different workflows?

Yes. Each workflow watching a column names the record types it takes, so a bug and a story reaching the same column can start different workflows.

Can several agents work on one board?

Yes. Each item is taken by one run at a time, so two agents never work the same item, and you set how many run at once. With the engine, every run gets its own working copy, and you can give bugs to Codex and features to Claude Code under the same rules.

Is this another tracker?

It's the board you know, extended for agents. Cards, columns and moving work by hand work as in any kanban. On top of that, every item is a record your agents read, a column can start a workflow, and agents pick up only ready work, in your roadmap's order.

Give your agents a board to work from

Define a board for one kind of work, and let its columns start your workflows.

Beta accessSee pricing