One connector · 20 tools
Objectives, key results, epics, meetings, docs and releases — all reachable over one authenticated connector. The board is where your team looks at what the assistant already did.
30 days free, no card. No AI credits, ever.
Things you can ask on day one. Every one of these maps to tools that exist.
The planning layer, not only the ticket layer — which is the difference between an assistant that files what you dictate and one that can tell you what is at risk.
Outcomes with numbers on them, checked in over time.
Tasks, bugs, chores and feedback, with legal status transitions.
A notetaker joins, and what was decided becomes work in the plan.
Specs, decisions and runbooks — written and diffed by the assistant.
What shipped, assembled from the work rather than typed again.
Reports arrive from a widget or SDK and become triaged work.
Most tools meter the AI: an allowance per seat, credits you buy, an agent that stops mid-sprint. Here the model is yours. Your assistant connects with your own subscription, we never call a model, and there is nothing to top up. The price is for the workspace.
An agent picking from a hundred near-identical tools picks wrong, and you find out when the ticket lands in the wrong project. The surface is twenty tools grouped by what they are about — work, docs, meetings, releases — each taking an action. Small enough to hold in your head is also small enough to be reliable in.
No custom fields, no custom statuses, no workflow builder. Objectives hold epics, epics hold work, and every type knows which transitions are legal. That is what lets an assistant operate the tool correctly without being taught your conventions first.
Tasks, bugs and chores share one board grouped by where the work is, not by what kind of thing it is — because a release contains both and a stand-up walks the columns once. The same data has seven chairs: an analyst opens on what is still unspecified, quality opens on what claims to be finished without meeting its own criteria, support opens on the inbox. Nobody is configuring a view; they are choosing a seat.
One price for the team, not per seat per month forever.
€19/month
One small team, getting the habit.
€49/month
The whole team, all the projects.
€149/month
Everything, plus what ships next.
No investors, no sales team, and no logo wall — you would notice, so it may as well be said here. What that buys you: decisions made in an afternoon, and a product with an opinion rather than a settlement between six of them.
What it costs you: fewer integrations than the incumbents, and a bet on something small. Self-hosting is supported precisely so that bet is reversible — the same container, on your own Postgres, with an export of everything on the way out.
Any client that speaks MCP. Claude Code and the Claude apps connect over an authenticated connector; anything else that implements the protocol works the same way. There is a read-only endpoint too, for a key you would rather could not change anything.
No, and there is nothing to top up. The assistant is your own subscription — we never call a model and never bill you for one. The price is for the workspace.
The workspace becomes read-only. Nothing is deleted and you are not locked out; everything is still there and still readable until you decide.
Yes. The board, docs and objectives are an ordinary web app, and it opens on a written status update rather than a wall of cards. The connector is how the work gets in quickly, not the only way in.
A notetaker joins the call from your calendar, records it and transcribes it. The transcript is searchable, the terms your team actually uses are learned rather than guessed, and the assistant turns what was decided into epics and work items linked to the objective they serve.
On EU infrastructure. Self-hosting is supported — it is the same container the hosted version runs, on Postgres you control.
No. Your content is not used for training and is not sold. The assistant that reads it is the one you connected.