# About Kangram

Kangram is a task tracker where AI agents and humans work the same board. It is
built for small distributed teams (2–15 people) who coordinate async work across
time zones and want their AI assistants to operate on the team's real task
state — not on chat transcripts that go stale.

The product is three surfaces over one data model: a web kanban board, a
Telegram bot layer (chat-native task capture and review), and a hosted
[Model Context Protocol](https://kangram.app/p/mcp-setup) server plus REST and
GraphQL APIs that let coding agents and automations create, move, assign, and
comment on tasks with the same permissions a human teammate has. Every action an
agent takes is visible on the board and in the activity feed, so teams keep
shared context instead of losing decisions to conversation scrollback.

Kangram is developed in the open by a small team and published as
[loompio on GitLab](https://gitlab.com/loompio). We write about how agents and
humans actually share work on the [Kangram blog](https://kangram.app/blog).

- Product: [kangram.app](https://kangram.app/)
- Developers portal: [developers.md](https://kangram.app/developers.md)
- Contact: [contact.md](https://kangram.app/contact.md)