New in Breeze: your AI assistant can now do the work, not just report on it

From asking to doing

Read-only was, honestly, a nicer dashboard. You could ask a question instead of building a filter, which saved a minute here and there. Useful, but the answer always ended the same way: now go into Breeze and do the thing.

Making changes is a different kind of help. The admin around your work — writing up tasks after a meeting, moving cards along, logging the hours you already worked, tidying a board nobody has touched in a month — is the part everyone puts off. It is also the part that is easiest to describe in a sentence and tedious to click through. That is exactly the work you can hand over now.

What this looks like in practice

Six things people are already doing with it. None of them need special phrasing — you type it the way you'd say it to a colleague.

Turn meeting notes into tasks

Paste your notes into the chat and say: "Turn these into tasks in the Q4 Launch project, put them in the To do column, assign the owners we named, and set them due next Friday."

Because that's more than one task, the assistant shows you the full list first — every task, its owner and its date — and waits. You read it, you say go, and the cards appear on the board. If one name was wrong you say so before anything is created.

Wrap up your day in one message

"Log 2 hours on the homepage redesign task for today, move it to Review, and comment that the copy still needs sign-off."

Three separate jobs — time tracking, a board move and a comment — in one sentence. The time lands on the right task with today's date, and your teammates see the comment in Breeze exactly as if you'd typed it there.

Clean up a board that got away from you

"Everything sitting in Done from last month — archive it."

Before anything disappears, the assistant tells you exactly what it's about to archive and how much goes with it. Then it waits for your yes. And archiving in Breeze isn't destruction: those tasks can be restored, so a cleanup you regret is a sentence away from being undone.

Start a project from a brief

"Set up a project called Website Relaunch with columns for Backlog, In progress, Review and Done, add Sarah and Tom, and create the first five tasks from this brief."

A new project, its board, its people and its opening tasks — from one message. It is the fastest version of the setup everyone does by hand on day one of a new engagement.

Turn a process into a checklist

"Add a Pre-launch checklist to this task with: proofread copy, check links, compress images, test the form, and set up redirects."

Useful when the process lives in your head, or in a doc nobody opens. Say it once and it becomes a real checklist on the card, ticked off as you go — and you can tell the assistant to tick items off later too.

Catch up, then act on it

"What changed in Marketing this week?" — then, once you've read it — "Reply on the newsletter card that we're pushing it to Monday, and change its due date."

This is the one that shows why reading and writing belong together. You look into the project, spot the thing that needs a decision, and make the decision without ever switching windows.

What it can change

In short, the everyday substance of a project:

• Tasks — create them, rename them, edit descriptions, set start and due dates, assign people, move them between columns, mark them done, archive and restore them.
• Checklists — add checklists to a task, add and rename items, tick them off, delete them.
• Comments and discussion — comment on a task, post in a project's discussion, edit or delete what it posted.
• Time and estimates — log time on a task, correct an entry, set and adjust estimates.
• Projects and columns — create projects, add and rename board columns, reorder them, add teammates to a project.

A handful of things are still hands-on: attachments, custom fields and creating new tags aren't part of it yet. For the exact, complete list — every action, with an example of what to say to trigger it — see the full list of what it can do in our help pages.

How we kept it safe

Letting an AI write into your project board is a bigger ask than letting it read. Here is what actually protects your data — not promises, but how it's built:

• It acts as you, with your permissions. It signs in with your Breeze account and obeys exactly the rules the app does. It can only change what you could change by hand. Observers stay read-only. Structural changes — new projects, new columns, adding people — still need to come from a team owner or admin.
• Nothing happens in bulk without you seeing it first. If it's about to create or change more than one thing, it stops, shows you the list, and waits for your go-ahead. Nothing is written until you say yes.
• Deletions always ask — even for one item. And the preview tells you what comes with it. Archiving a column tells you how many tasks go too; removing a project tells you how many tasks and comments that means. A small number can never hide a big one.
• Most things come back. Archived tasks, columns and projects can be restored, from the assistant or from Breeze. Comments, checklist items, time entries, estimates and discussion posts are deleted for good — which is why those are the ones it's most careful about.
• A runaway assistant can't run away far. Each connection has a ceiling on how much it can change in an hour, plus a limit on firing the same action over and over in quick succession. If something goes wrong in a loop, it stops early rather than rewriting your board.

Already connected? Reconnect once

If you set up Breeze in your AI assistant before today, that connection was granted read-only access — so it will politely refuse to change anything until you reconnect. It's a one-minute fix: remove Breeze in your assistant's connector settings and add it again at mcp.breeze.pm/mcp. The connect screen shows you what's being granted before you approve it.

Haven't connected yet? The setup steps for Claude, ChatGPT and other assistants take about a minute, and it's a one-time thing.

A few honest limits

• @mentions don't notify yet. If the assistant writes a comment mentioning someone, the mention shows up as text but doesn't send them a notification. Mention people yourself when it matters.
• Brand-new items take a moment to become searchable by meaning. Keyword search finds them immediately; the AI search index catches up shortly after.
• Check anything that has money or a deadline attached. The assistant is good, not infallible. Billable hours and hard dates deserve a glance in Breeze itself.

Less clicking, same board

Nothing about how Breeze works has changed. Your board is your board, your team sees the same thing they always did, and every change the assistant makes shows up in Breeze under your name, in the activity feed, like any other change.

It also sits alongside the AI already built into Breeze — asking questions inside the app still works exactly as before. This is the same idea, moved into whichever assistant you already have open all day.

What's different is that the boring half of project management — the writing-up, the moving, the logging, the tidying — is now something you can describe instead of do. Connect your assistant, ask it to clear out last month's Done column, and see how it feels. The help page has the setup steps and everything it can do.

Common questions

Can my AI assistant change my Breeze data now?

Yes. It can create, update, move, complete and delete work in Breeze — tasks, checklists, comments, time entries, estimates, projects and board columns — on top of answering questions the way it always could.

Could it delete something by accident?

It asks before every deletion, even a single one, and shows you what goes with it first. Archived tasks, columns and projects can be restored. Comments, checklist items, time entries and estimates are permanent once deleted, so those are confirmed most carefully.

Do I need to reconnect if I set it up before?

Yes. Connections made before this release were granted read-only access, so they cannot make changes. Remove Breeze in your assistant's connector settings and add it again at mcp.breeze.pm/mcp to grant the new permission.

Can it change things I don't have access to?

No. It signs in as you and follows the same permissions as the Breeze app — your team, your projects, your role. Observers stay read-only, and structural changes like creating projects or adding members require a team owner or admin.

Which AI assistants does this work with?

Claude, ChatGPT, and any other assistant that supports custom connectors. You connect once using the address mcp.breeze.pm/mcp and your normal Breeze login.