Clawdeck got the problem right. lst.so gets the solution right.
In 2024, I built Clawdeck: an open source kanban board for managing AI agents. It picked up hundreds of GitHub stars, thousands of users, and real traction in the OpenClaw community. People were hungry for it.
On May 30, 2025, the hosted version of Clawdeck.io is shutting down.
This post is about why, and what I learned building it.
What Clawdeck got right
Clawdeck was built on a real insight: AI agents need somewhere to pick up work and report back. Treating agents as workers who need assignments, not just chat windows you prompt and forget, was the right frame.
That framing is still correct. It is the foundation lst.so is built on.
The problem Clawdeck identified, that you need a shared layer between humans and agents for work to happen reliably, is exactly the problem worth solving. Thousands of developers agreed. That was not the mistake.
What Clawdeck got wrong
Clawdeck modeled agents as separate workers who operated in their own space. You would assign a task, the agent would run, and eventually something would come back. Two worlds, occasionally syncing.
The visibility gap was brutal. Agents were doing things I could not see. I could not see what they were doing. Every session started with re-explaining context. Every handoff between agents required manually bridging the gap. The overhead compounded: server maintenance, overnight debugging, token burn while agents sat idle.
I stopped using Clawdeck. Not because the problem went away, but because a separate workspace for agents made the fragmentation worse, not better. Now there were two systems instead of one.
The insight: agents do not need their own workspace. They need to share yours.
One list. Shared by you and every agent you work with.
When you add a task from your phone, your agent sees it next time it connects. When Claude Code finishes a subtask, it checks it off and logs what it did. When you open lst.so, everything is there: your tasks, agent tasks, what got done overnight, what is still open. No re-explaining. No bridging. No second system.
Every action is logged with who did it, you or the agent. The activity log becomes the shared memory that survives across sessions.
This is lst.so.
How it works
lst.so has a full MCP server published as lst-mcp on npm. Connect any agent in about ten minutes.
{
"mcpServers": {
"lst": {
"command": "npx",
"args": ["-y", "lst-mcp"],
"env": {
"LST_API_KEY": "your-api-key"
}
}
}
}
Works with Claude Code, Cursor, ChatGPT, and anything that speaks MCP or HTTP. Once connected, agents can list tasks, create new ones, log progress, and check off subtasks, all in the same list you are working from.
If you used Clawdeck
The open source repo stays up. Self-hosted installs keep running. Nothing changes if you run it yourself.
If you have data in Clawdeck.io, export it before May 30 by giving your agent this prompt:
"Log into clawdeck.io with my credentials and export all my boards, columns, and cards to a local JSON file."
Then give lst.so a try. The instinct that brought you to Clawdeck is exactly what lst.so is built around. Same problem, better solution.
Frequently asked questions
Does lst.so work with Claude Code?
Yes. Install lst-mcp via npm, add the config above with your API key, and Claude Code can read your task list, create tasks, log progress, and check off subtasks without leaving the terminal.
Does it work with Cursor or ChatGPT?
Cursor supports MCP natively, same config works. ChatGPT supports MCP through the desktop app. Any agent that speaks HTTP can use the REST API directly.
How is lst.so different from Linear or Jira?
Linear and Jira are built for human engineering teams. lst.so is built for humans who delegate work to AI agents. Every task has an agent field. The activity log tracks who did what, human or AI. Agents connect via MCP to pick up and complete work autonomously.
What does it cost?
$9/month. One plan, 7-day free trial, no credit card required to start.
Can I import from Clawdeck?
Not directly. Export your boards to JSON before May 30 and recreate what you need in lst.so.
About the author
Andres Max is a software engineer and product builder who has been building AI-native tools since 2023. He created Clawdeck and built lst.so out of the lessons it taught. He writes at mx.works.