Your backlog,
their execution

When you drop a task into the In Progress column, AgentsRoom creates a temporary agent with the task title as its initial prompt. If you assigned a specific agent role to the task, that role's system prompt is used. The agent's terminal opens automatically so you can watch it work.

Yes. Drag as many tasks as you want to In Progress. Each one spawns its own agent with its own terminal. You can run five, ten, or twenty tasks at the same time, limited only by your machine and your provider's rate limits.

The task stays in In Progress. You can give new instructions to the agent directly from its terminal. When you are happy with the result, drag the task to Done. You can also move it to Pending if you need to wait for something else before continuing.

Yes. Tasks are synced to the AgentsRoom cloud and linked to your project via a stable backlog ID stored in .agentsroom/project.json. Any device with the same project sees the same backlog.

Yes. Each task supports image attachments. Add mockups, screenshots, error captures, or design references directly to the task card. These are uploaded and stored securely.

Yes. Select several tickets on the board, or use the card menu, and merge them into one. The AI recombines their titles and descriptions into a draft you edit before confirming, upvotes add up, and comments, screenshots and links move to the survivor. The rule that matters: no reporter is ever lost. Everyone who reported an absorbed ticket stays subscribed to the survivor and is notified when the fix ships, and the public links already sent to them redirect to it.

Yes. A ticket can pin its own provider, model, reasoning effort, account profile and skills, and those overrides win over the default agent profile when the ticket spawns its agent. Columns can also be sorted on demand, on desktop and on mobile, and agents can create, update, comment, reorder and delete tickets over MCP without leaving the terminal.

Yes. Link two tickets as blocks and blocked by, and the board enforces the link instead of just documenting it: while the blocker is still open, the blocked ticket sits in Pending rather than TODO, so Start next never picks it up. It returns to TODO on its own when the last blocker closes, and goes back to Pending if that blocker is reopened. A ticket already in review or done is never parked, and a ticket you parked by hand stays where you put it. Agents handle the links themselves over MCP: one that discovers a prerequisite mid-run creates that ticket and links it in a single call, and a ticket that turns out to be blocked while in progress is moved out of the way without killing the terminal running it.

In the idea box, a board attached to your account rather than to a project. Open the backlog board of any project, click the project name at the top and pick Idea box. Tickets there work like the others: title, description, tags, screenshots, synced on your account and visible from the mobile app. The one thing the idea box cannot do is start an agent, since there is no project to run it in: starting an idea opens the move dialog, and the ticket runs on the project you pick.

Yes. Open the card menu, or select several tickets, and choose Move to another board. The ticket keeps its description, its attachments and its history and lands at the top of its column on the target board, from the desktop or from the ticket detail on the phone. Two things are left behind on purpose: the agent assigned on this machine, and the links to other tickets, because a dependency between two boards could not be seen from either. A ticket reported by a client through the public backlog never moves: its public page and its conversation thread belong to the board it was submitted to.

Yes, since AgentsRoom 1.194.0. Every save that changes a ticket's text or configuration (title, description, type, tags, private note, assigned agent or team, skills, provider, model, reasoning effort, account profile, worktree options, reviewer) keeps the previous version, including a change made by an agent through MCP. On the desktop, open History from the ticket header to compare any earlier version with the current one and restore it; on the phone the version is loaded into the ticket form as a draft and saved normally. Board state is not versioned: dragging a card, changing its priority or marking it won't fix creates no version. The same history now exists for scheduled agents and for memory notes (memory on the desktop only), next to the one prompts and skills already had.

Yes. Every ticket has a Comments section, the team's internal thread: you write in it from the ticket, your agents through the AgentsRoom MCP, to log progress, a hand-off, a blocker or a question. It is never shown to the person who reported the ticket and never emailed. The card shows how many comments a ticket has, and an agent started on the ticket reads the latest ones before it begins. In the project settings you can name a lead agent, a role such as the product manager or one saved agent: each time another agent comments, the lead gets a message so it can answer on the ticket, reassign it or bring you in.

On the card. A ticket set to run in a separate worktree shows its branch name at the bottom of the card, on the desktop and on the phone, and hovering it gives the full branch and whether it merges back on its own. Two tickets planned in parallel no longer look like two tickets sharing the same checkout.