Cursor Kanban Agent: Drag a Task, Get a Pull Request
Malik Farooq
May 3, 2026
Deep Dive
Cursor Kanban Agent: Drag a Task, Get a Pull Request
Introduction: The Kanban Board That Codes Itself
On April 29, 2026, Anysphere — the company behind Cursor — published a project called agent-kanban in the official cookbook for their new TypeScript SDK. The concept is deceptively simple: drag a card from one column to another on a Linear-style board, and an AI agent in the cloud takes over, writes the code in an isolated virtual machine, opens a pull request on GitHub, and attaches the result as an artifact on the card itself.
It's not a standalone product. It's a working example of what the Cursor TypeScript SDK makes possible.
What the Cursor Kanban Agent Actually Is
agent-kanban| Project | What it does |
|---|---|
| Drag-and-drop board that triggers AI coding agents |
| Automatically reviews PRs with agentic feedback |
| Labels and prioritizes GitHub issues autonomously |
| Generates documentation from code changes |
The kanban project is the most visually striking — it makes the agent's work visible in real time as cards move across columns.
How the Cursor TypeScript SDK Works
The SDK exposes three core primitives:
- Sandbox VMs — Isolated cloud environments where the agent executes code safely.
- Subagents — Specialized agents that can be spawned and coordinated from a parent agent.
- Hooks — Event triggers (like a card drag) that launch agent workflows.
This architecture enables the kanban board: a card drag fires a hook, which spawns a subagent in a sandbox VM, which runs the coding task and pushes the result.
The Cursor 3 Context: Agents Window and Token Pricing
The TypeScript SDK is part of the Cursor 3 release, which introduced the Agents Window — a dedicated panel for monitoring and managing cloud agents running in parallel. Token-based pricing applies: you're billed per agent-compute-minute, not per message.
Practical Use Cases for Development Teams
- Sprint planning to code in one drag — Move a Jira-style ticket to "In Progress" and get a PR by EOD.
- Bug triage automation — Move a bug report to "Assigned" and let the agent attempt a fix.
- Feature scaffolding — Drag a feature card to "Build" and get boilerplate + tests generated automatically.
Conceptual Representation
In Practice
Conclusion
The Cursor Kanban Agent is the clearest demonstration yet of what agentic development infrastructure looks like in practice. It's not hype — it's a working SDK example you can clone and deploy today. For teams that want to experiment with AI-native development workflows, this is the most tangible starting point available in 2026.
References
[1] Cursor TypeScript SDK Cookbook, Anysphere, 2026.
[2] Cursor 3 Release Notes, 2026.
[3] Internal Latest AI Team Research, 2026.
[4] Original Article: https://pasqualepillitteri.it/en/news/1717/cursor-kanban-agent-sdk-2026
Ready to master AI?
Join 1,000+ professionals getting the edge in AI marketing. 3 minutes a day to 10x your growth.
Join Free NowKeep reading
Meta Ads MCP for Claude
Learn the latest AI strategies to stay ahead in the marketing game.
Malik Farooq/
NotebookLM April 2026: Mobile, Cinematic Video & Gemini Sync
Learn the latest AI strategies to stay ahead in the marketing game.
Malik Farooq/
NotebookLM April 2026 Update
Learn the latest AI strategies to stay ahead in the marketing game.
Malik Farooq/
Back to archive
Share