AI that knows the board, not just the buzzwords
Tarsen’s AI features are useful because they are attached to the cards, projects, permissions, and activity already in the workspace. The core board still works without AI.
Three layers, with different jobs
| Layer | What it does |
|---|---|
| Mission Mode | Ranks open work into a capacity-aware daily plan; its core scoring is deterministic. |
| Ask Tarsen | Answers questions about the accessible workspace and proposes card changes for confirmation. |
| MCP server | Lets a compatible assistant read or change the real board through an API key and the user’s permissions. |
The assistant does not get a blank cheque
- Ask Tarsen can propose, but writes inside the app require confirmation.
- Every proposal is re-validated against the user’s current project access on the server.
- MCP keys are scoped to one workspace and act with the permissions of the connected user.
- Created cards, updated cards, and comments appear in the normal activity stream.
Useful questions to ask
- What is overdue this week, and what is blocking it?
- What changed on the roadmap while I was away?
- Summarize the comments on TAR-142 and tell me what was decided.
- Create a card for the billing migration with a Friday due date.
The answers come from the board the user can access. The assistant is not a replacement for permissions, and it cannot infer private projects it cannot see.
When to leave AI switched off
If your team does not want a model provider in the loop, keep it off. The Kanban board, search, keyboard layer, permissions, imports, exports, and deterministic Mission Mode remain useful without an AI key.
Something missing here? Email support@tarsen.co and we’ll write it.