# Kangram — task management where AI agents and humans work the same board

> Markdown twin of the [Kangram homepage](https://kangram.app/) for agents and
> crawlers that prefer text. Served at `/index.md`; negotiated on `/` when the
> client sends `Accept: text/markdown`.

Kangram is a task tracker for small distributed teams (2–15 people): a web
kanban, a Telegram bot, and a hosted MCP server, so AI agents and teammates
create, move, and comment on the same tasks. Agents connect through MCP or the
REST/GraphQL API; humans review everything from the board or the team chat.

## What teams use it for

- Turning conversations into tracked tasks — from Telegram chats or AI coding sessions.
- Shared visibility: overdue, blocked, and unassigned work surfaces on every board.
- Cross-session agent memory: handoffs keep context between AI work sessions on a board.

## For developers and agents

- [REST API docs (OpenAPI)](https://kangram.app/api/v1/docs) · [OpenAPI schema](https://kangram.app/api/v1/docs/openapi.json)
- [GraphQL schema (SDL)](https://kangram.app/api/v1/docs/graphql.sdl) · endpoint [https://kangram.app/graphql](https://kangram.app/graphql)
- [MCP server card](https://kangram.app/mcp/server-card) · [Agent guide](https://kangram.app/agents.md) · [llms.txt](https://kangram.app/llms.txt)
- [Developers portal](https://kangram.app/developers.md) · [Pricing (machine-readable)](https://kangram.app/pricing.md) · [MCP setup guide](https://kangram.app/p/mcp-setup)