Source: https://harnsy.dev/features/tasks

> A task board for Claude Code, Codex, OpenCode and pi agents: types, subtasks by role, checklists, reports and screenshots, and your accept or return.

# From your request to your “Accept”.

The lead turns what you ask into tasks on the board, splits them by role and brings back the result with proof. You decide on the task card.

The task board: one card waits for your acceptance; a backlog card moves only when you send it to work. Sample data.

### A request becomes a task.

Say what you need; the lead records it as a task with acceptance criteria and a type: bug, feature, research or task. The board filters by type.

### Split by role.

The lead splits a task into sub-tasks for the roles below it. The parent waits until every sub-task is settled; a sub-task in review shows on the board as “part of #N”.

### Priority.

Mark a task high, medium or low. The backlog sorts by priority or by date, so the work that matters most is on top.

### Tasks that wait for each other.

Say which task blocks which, also across projects. A card names its blockers, and says “Ready to start” when nothing holds it. A dependency is the order of work: it shows, it does not forbid. When a blocker is done or dropped, the blocked task’s agent and lead are told.

### Chains.

Tasks › Chains draws linked tasks as chains, step by step from left to right (from top to bottom on a phone). In a task, Dependencies shows its whole chain: what comes before it and what waits for it.

The Chains view: linked tasks drawn as chains, one step after another. Sample data.

### Every task says what it waits for.

Cards on the board, the task dialog and sub-items show “waits for … · who · how long”: an agent, the lead, you, a review or something outside. A task you block says whether it waits for you or for something external.

### Defer a task until a date.

“Defer until…” puts a task away until a date; on that day it returns by itself and its lead is told. A task in review can be snoozed the same way: it stays in review but does not wait for you until then. The “Deferred” column is folded by default, soonest first.

### A backlog for later.

Park what can wait: nobody takes it, and it doesn’t count as open work. Send it to work when its time comes, or reject it with a reason.

### Group tasks into releases.

Assign tasks to a milestone, a release of your project with a due date and a theme, filter the board by it and watch the progress strip and the burn chart. Everyone sees milestones; a harnsy Max key creates them.

### Progress you can see.

Agents split a task into checks and close them as they go; the card shows a progress bar. The history is folded by default and reads as short lines: who did what, and when. Criteria, results and the history render Markdown: tables, lists and links.

### Branch and folder on the card.

A card shows the git branch and worktree its agent works in, with “copy path”. When the task closes, the agent is reminded to remove the worktree.

### Work comes with proof.

Agents hand in work with Markdown reports, screenshots and HTML pages, such as a chart or a table, shown in a sandboxed frame.

### You accept or return.

A finished task waits for you in review. Accept it, or return it with a comment that goes straight to the lead’s session.

### Statistics.

Done, dropped and returned tasks by day, for today, 7 and 30 days, who closed them, and GitHub issues filed from cards.

### All projects.

A summary for every project and in total: what is open, in work, waiting for you, stale and in the backlog, and what was created, done and dropped today, over 7 and over 30 days. Each project keeps its own board view: columns, filters and sub-items. Free harnsy runs one organization with 2 projects; a harnsy Max key adds more.

### Pick one of several options.

An agent can submit 2 to 5 options with pictures. You press “Pick Vn”, and you have 10 seconds to undo. A “choose” request shows its options as buttons, too.

The task board filtered by one milestone. Sample data.

The agents write code faster than I can read it.

## In harnsy you don't read every line: the task moves from role to role, and you accept the finished result.

In review you don't just click Accept or Return: write what to fix, and your comment lands in the lead's session right away. The lead sends the task back to whoever did it, with your words.

Howthe lead assigns it, each role hands it on with evidence, and the finished card turns amber. Changes shows every worktree’s diff when you want to look.

waits for you

Illustrative demo · #42: analyst → developer → tester → your acceptance.

A task with three options: pick one, or return it with a comment.

## Limits, honestly

- Attachments: Markdown and HTML up to 512 KB, images up to 5 MB, up to 50 per task.
- Checklists are read-only in the dashboard: agents close the checks.
- Tasks move through the team under the lead’s direction, not by a fixed process engine.

## Your first team is one prompt away.

Installing takes one sentence to your agent. To run your agents, harnsy drives a terminal: tmux if it is installed, otherwise its own terminal host (on Windows, always its own).

`Install harnsy following https://harnsy.dev/llms.txt`
