Most AI coding tools drop you into a chat. That is fine for one task, but it does not scale to a queue of work. Lanes uses an issue board instead, so agent work has structure: a backlog, work in progress, and things waiting on review.
How Lanes does it
- Columns for the full loop. Plan, implement, review, done. The board runs six columns in total: Backlog, Planning, Implementation, Review, Done, and Misc as a catch-all. Agents started in plan mode land in Planning; agents changing code live in Implementation.
- Live sessions on cards. Each card can run an agent with its own worktree and terminal. Click a card and the detail panel opens with the terminal on the left and the issue's notes, instructions, and metrics on the right.
- Status at a glance. Busy, awaiting input, stopped, or error on every card, with a bell when one needs you.
- Labels and filters. Slice a large board down to what matters right now. Labels are per-repository and color-coded, they combine with directory and workflow-step filters, and a filter badge reminds you the board is not showing everything.
Issues are cheap to create. Press Cmd+N, give the task a title and instructions, and the instructions become the agent's initial prompt when a session starts. Deep links go one step further: a lanes://new URL from Linear, GitHub, a script, or a bookmarklet opens Lanes with a draft issue already filled in.
Order is explicit too. Dependencies link issues that block each other, and cycle detection keeps the graph a DAG. Blocked issues carry a badge on the board, and the queue manager skips them when it auto-fills session slots from the backlog, so chained work runs in the right order without you sequencing it by hand.
At scale the board has the bulk tools you would expect: multi-select with Shift+Click or Cmd+Click, a bulk toolbar for stop, remove, and complete, board tabs for all issues, per-project, and per-worktree views, and sorting by manual order, newest, updated, or title.
Why it matters
A board gives you oversight. You see the whole pipeline instead of one chat window, and you decide what each agent picks up next. Planning, running, and reviewing live in the same place.
It also gives agent work a memory. A chat transcript scrolls away; a card keeps the task, the instructions, the session output, and the cost in one place you can come back to tomorrow.
Set it up
- Add a project and press Cmd+N to create your first issues. A title plus instructions is enough.
- Label as you go. Right-click a card, choose Add Label, and build the taxonomy inline.
- Wire dependencies for work that must land in order: right-click, Add Dependency, pick the prerequisites.
- Start sessions with Plan or Implement, and lean on board tabs and filters once the card count grows.
Further reading
- The issue board: columns, multi-select, board tabs, and sorting.
- Labels and filtering: per-repo labels and how filters stack.
- Dependencies: blocking, cycle detection, and queue automation.
- Deep links: pre-filled issue drafts from other tools.