One backlog for humans and AI agents
Kangram is a task tracker with a first-class Model Context Protocol server and a Telegram bot. Your coding agents create and track work in the same place your team does — no copy-paste, no context loss.
How it fits your workflow
Three surfaces, one source of truth. Agents talk MCP; your team talks Telegram or web.
Connect your agent
Add the Kangram MCP server to Claude, Cursor, or any MCP client. One config block.
Agents create tasks
Ask in natural language — tasks land on a shared board with status, priority, and assignee.
Team picks up in chat
Teammates see and act on those tasks in Telegram. One backlog, no sync overhead.
Works where you already work
Any MCP-compatible client, plus the chat app your distributed team lives in.
Claude Code
Create, list, and move tasks mid-session.
Cursor
Track TODOs on a shared board without leaving the editor.
OpenCode
Hand work off to a persistent task backlog.
Codex
Plan sprints and assign work to teammates.
Windsurf
Keep agent context across long runs.
Telegram
Manage everything from a chat — no new app to learn.
Built for real task management
Not a toy TODO list — a tracker teams can run on.
Boards & tasks
Kanban boards, tasks, subtasks, checklists, comments, priorities.
MCP-native
A first-class Model Context Protocol server your agents call directly.
Telegram-first
Built around chat — the place async teams already live.
UCL
One command layer powers MCP, REST, CLI, and the bot. No drift.
Start in under a minute
Add one block to your MCP client config and your agent has a task board.