2026-11-03
Your AI assistant now works on the project map
Depensee now ships an MCP server: Cursor, Claude or a company's own model can read boards, create tasks and lay them out on your command. What it is and why it matters for a business.
- #product
- #ai
- #integrations
The “LLM integration” item on our roadmap is done: Depensee now has an MCP server.
What it is
MCP (Model Context Protocol) is an open standard for connecting AI assistants to outside services. Cursor, Claude, VS Code, Codex and corporate chat platforms support it. Connect Depensee, and the assistant sees your boards the way you do: tasks, statuses, dates and the links between them.
What it gives you
- Decomposition in seconds. “Split this task into subtasks in the order they need to happen” – and the cards appear on the map, already linked with “contains” and “blocks”. People review instead of typing from scratch.
- Answers about the project without manual digging. “What is blocking the release?”, “which tasks have no due date?” – the assistant reads the dependency graph and answers to the point.
- Routine in plain words. Change statuses, set dates, link tasks, tidy the board – one sentence in the chat.
- The team sees it at once. The assistant’s edits reach every open map in real time, just like a teammate’s.
Under control
The assistant uses a personal token from the account settings and gets exactly its owner’s rights: a viewer does not become an editor. A token can be read-only, can expire, and can be revoked in one click. By design, the assistant cannot delete tasks.
The model does not have to be in the cloud: one running inside the company network works too, as long as it supports tool calling.
Try it at demo.depensee.com: Settings → “AI agents (MCP)” → “Create token”; the ready-made client config is right there.