Glossary
The words the board and these docs use, in alphabetical order. Each entry ends with the page that shows it in use.
The AI that works on a ticket: it reads your brief, proposes a plan, does the work and asks you when it needs a decision. One agent runs per ticket. See What an agent sees and can do.
App Space
Section titled “App Space”A gallery of single page tools, opened from More actions > App Space in the header. You fill it by saving an HTML artifact as an app, or by uploading a page; an app runs sandboxed, with no access to the board. See Find and reuse what agents made.
Artifact
Section titled “Artifact”A file or a link an agent registers on its ticket as a deliverable: a report, a chart, a web page, a 3D model. The board keeps its own copy of a file, previews it on the ticket, and lists every artifact of every ticket in the Artifacts space. See Find and reuse what agents made.
The kanban board: the columns from Backlog to Discarded, and the tickets in them. One instance can hold several boards, each with its own working directory and overrides. See Organize the board.
Catch up
Section titled “Catch up”A card by card walk through every ticket with news since you last looked, answered with a swipe or an arrow key. See Choose how the board notifies you.
A conversation with an agent, started in Chats, for a question rather than a piece of work. It is a ticket underneath, off the board until you add it. See Ask in Chats instead of a ticket.
The Getting started coach: after the setup wizard, it rings one control at a time (point the board at your code, create a ticket, approve a plan, merge) until you have done each once. See Choose a profile, and change it later.
Command ticket
Section titled “Command ticket”A ticket that runs one skill or slash command, such as a repo report, instead of a free text brief. See Run a repo report.
Engine
Section titled “Engine”The agent CLI a ticket runs on: Claude Code (the default), Codex or Cursor. Engines differ in what the board can enforce on them. See Run Codex or Cursor instead of Claude Code.
A ticket no agent runs: it groups subtickets under one outcome. See Split big work into subtickets.
Flow plan
Section titled “Flow plan”A plan drawn as a graph of blocks, each a behaviour, which you edit and comment on. One of the three plan types. See Review and annotate a plan.
A Loop that runs again and again until its objective is met or its budget is spent, keeping a ledger of what each run did. See Run a Goal until it is done.
A how-to page in these docs. A guide whose steps happen on a control also carries a Show me walk through that rings each control on the board in turn. See Your first finished ticket.
In Review
Section titled “In Review”The column where finished work waits for your OK: read what the agent made, ask for changes, or approve it. See Review and merge an agent’s work.
Interactive plan
Section titled “Interactive plan”A plan written as an interactive web page, which you annotate by pointing at any element. One of the three plan types. See Review and annotate a plan.
Knowledge Base
Section titled “Knowledge Base”The board’s wiki of pages and links, grouped in spaces and folders. Agents search it before they guess and propose a page when they learn something durable; you promote what stays. See Keep what agents learn.
A ticket that runs work on a schedule or on an event: each firing creates a run, which executes the Loop’s prompt or its pipeline. See Run a Loop.
One step of a Loop’s pipeline: a source that reads the outside world, an agent, a script, a subticket, a gate, a sink that writes a result. See Loop nodes.
Orchestrator
Section titled “Orchestrator”An agent that splits a ticket into independent strands, dispatches them as subtickets (usually through a Loop pipeline), and then merges and reviews what they bring back. See Split big work into subtickets.
What a ticket does when it needs something only a human (or its orchestrator) can give: its run stops, it moves to Pending Action, and its session waits. Your answer resumes the same session, so nothing is lost. See Answer what the board asks you.
Pending Action
Section titled “Pending Action”The column of tickets that wait on you: a plan to review, a question, a tool request, a manual step, a merge conflict or a spent budget. See Answer what the board asks you.
Permission gate
Section titled “Permission gate”The check every tool call and shell command of an agent passes through: allowed, denied, or routed to you as a tool request. See The permission gate.
Pipeline
Section titled “Pipeline”The graph of nodes a Loop runs, edited in the Loop’s full screen editor. Its edges are the order the nodes run in. See Run a Loop.
A ticket’s proposal, written before any change, that you annotate, send back or accept. Also the name of the plan type written as Markdown text; the other two are the flow plan and the interactive plan. See Review and annotate a plan.
Profile
Section titled “Profile”The starting set of settings the setup wizard writes: Explorer, Developer or Maintainer. Every value stays editable afterwards. See Choose a profile, and change it later.
Quality gate
Section titled “Quality gate”The code quality rules agents are held to (per language: file length, complexity, coverage) and the check that stops a ticket reaching In Review while they are broken. See Hold agents to a quality bar.
Ratchet
Section titled “Ratchet”How the quality gate judges by default: only code your change makes worse fails, so an edit to an old file is not blocked by the debt already in it. See Hold agents to a quality bar.
Review lane
Section titled “Review lane”The review agent: when a ticket reaches In Review, it can review the branch on request or by itself and post what it finds. Off unless you turn it on. See Review and merge an agent’s work.
One stretch of an agent’s work on a ticket, from the moment the runner starts it until it finishes, asks or parks. A ticket often has several runs over one session. Also: each execution of a Loop, which is a ticket of its own. See What an agent sees and can do.
Scratch workspace
Section titled “Scratch workspace”The private folder a ticket works in when it names no repository: no branch and nothing to merge, the right home for research and one off files. See Worktrees and branches.
Session
Section titled “Session”The agent CLI’s conversation for a ticket, saved by the board. A reply, an approval or a restart resumes the same session, with everything the agent already knew. See What an agent sees and can do.
Subticket
Section titled “Subticket”A ticket under a parent ticket, with its own agent, branch and review. You create them by hand, from a plan, or an orchestrator dispatches them. See Split big work into subtickets.
Ticket
Section titled “Ticket”One unit of work for an agent: a title and a brief, a mode, a working directory, and a timeline of everything that happened to it. See Write a ticket an agent can finish.
Verification check
Section titled “Verification check”A command agreed with a plan (a test suite, a build) that runs before the ticket settles in In Review; a failure sends the work back to the agent to fix, a few times at most before it parks for you. See Hold agents to a quality bar.
Worktree
Section titled “Worktree”A separate checkout of a repository on its own branch, one per ticket, so agents on the same repository never step on each other. See Worktrees and branches.