A board inside every card
A Story is one unit of work. Open it and you get another kanban board, with the same four columns, whose cards are that Story's Tasks. An agent moves Tasks the way you drag them.
Idea
✨Offline board cache
🚧Split staging environment
Planned
✨Saved filters
🪲Tag filter drops on reload
In Progress
🪲Tag filter drops on reload
Done
🧹Drop the legacy importer
🪲Tag filter drops on reload
Write a Story, break it into Tasks, and move it to In Progress. Moving the card is the hand-off.
It reads the board through MCP, picks up the Story that is In Progress, and moves each Task along as the work lands. Your open tab updates as it goes.
The Story moves to Done, and the agent writes what actually shipped into its description. After 30 days in Done it is filed to the Changelog.
The board
A Story is one unit of work. Open it and you get another kanban board, with the same four columns, whose cards are that Story's Tasks. An agent moves Tasks the way you drag them.
Every agent write checks whether the Story changed since the agent read it. If you both changed the same part, the write is refused and the refusal shows both versions.
Whenever anything on a board is written, by you or by the agent, every open tab updates. You watch the work move across the board while it happens.
Anything an agent creates, edits, or merges gets the 🤖 ai-generated tag. A Story's Activity Log lists its recent writes, newest first, each marked with how it arrived.
Your board is private until you choose to make it public. Nobody else can open it or tell that it exists, and only you can change it, whether from the browser or through your agent.
Collapse the lanes and the board's dashboard takes their place. Needs attention names the Stories that are overdue, due in the next week, waiting on a Story that isn't done, or marked Done with Tasks still open.
The Changelog shows a year of shipped Stories, one square per day. Stories carry due dates, tags, and Markdown descriptions, set on the board or through MCP.
The method
The board and MCP let an agent edit a plan. The skills are what make tada a way of working: two loops, each one feeding the board.
The board drives the work, and the work updates the board.
What the audit finds becomes Stories on the board.
Set up
Step 1
Boards are hosted, so there is nothing to install to use one. Sign in and you land on your board.
Open your boardStep 2
Add it to the project's .mcp.json. The agent working in that project
gets a set of plan tools for that board.
{
"mcpServers": {
"tada": {
"command": "npx",
"args": [
"tsx", "/path/to/tada/app/mcp/server.ts",
"--project", "myproject",
"--owner-email", "you@example.com",
"--url", "https://app.tada-ai.app"
]
}
}
}
Boards are created for you today, not self-serve. There is no “new board” button yet, and the ready-to-paste config comes with your board. Read the quick start
Questions
The docs cover the rest.
Not yet. Boards are created for you today, and there is no “new board” button. Once you have one, everything on this page works.
tada ships a stdio MCP server, so it is for a coding agent that can load one. The
quick start shows the setup for Claude Code, which reads .mcp.json from
the project root.
If you changed different parts, the agent's write goes through, and its reply notes that someone else was editing. If you both changed the same part, the agent's write is refused and the refusal shows both versions. Neither edit is silently lost.
A board is private until you make it public. Nobody else can open a private board, or tell that it exists. Anyone can read a public board, signed in or not. The privacy page has the detail.
tada takes no payments. It has no ads and sells no data.
No. This site runs no analytics and no scripts. The app counts page views with Vercel Web Analytics, which uses no cookies.
Open your board, then give your agent the same one.