Crewmate Profiles, Dynamic Dashboards, and Remote Crews as Live Windows
0.9.0Kiro Crew 0.9.0 lets you check on a crewmate at a glance. Opening one shows where its work stands, and a Profile card holds its permission, model, and schedules. In preview, each crewmate gets a dashboard page built from its own record, and remote crews open as live windows onto the other machine. The composer warns before a prompt outgrows the model, the sidebar hides and tidies what you do not need, and the Usage tab estimates token use and projects when your credits may run out.
These changes affect new installs, saved settings, or where controls live:
- New installs answer more briefly. A new install starts Settings → Chat → Transcript → Response Verbosity at Answer only. An existing config keeps its stored level.
- The Ultra verbosity level is retired. A config that stored Ultra now behaves as Answer only. Choose another level under Response Verbosity if you want longer replies.
- Translucent panels are now off by default. The top bar, message box, and search fields render as solid cards. Turn on Settings → Display → View → Translucent panels to bring back the frosted glass that 0.8 shipped on.
- Old relay chats open read-only. A chat that an older build ran on a peer crew through the retired turn relay still opens, but you cannot send to it.
- Settings moved. Agent Capabilities is now labeled Customize. Subagent Settings and Warm Pool now live under Settings → Agent Harness, and the Warm Pool card is labeled Pre-started Sessions.
Opening a crewmate on the Crewmates page shows a card with its goal, what is finished, running, or waiting on you, and one next step. The card reads from the crewmate's record, so it costs no chat turn. When no goal is in flight and the thread is new or idle for six hours or more, the card instead recaps the goals the crewmate left open and its recent sessions.
The pill at the top of a crewmate's thread says what it is doing now. Click it to open a Profile card with the crewmate's memory, workspace, and notes, its own Permissions, Model, and Reasoning effort, a Schedules tab for what wakes it, and a Sessions tab for the work it drives.
With a thread open, the roster folds into a switcher of faces in the thread header, and Show the full roster brings the column back. By default the roster lists crewmates you have chatted with or starred. The open crewmate stays listed while search is empty, and typing a search reaches the rest.
Turn on Settings → Developer → Feature Previews → Dynamic Dashboard and a crewmate's side panel opens a Dashboard tab that shows a project report by default. Twelve pages ship with the product. Ask the crewmate in chat for another one, such as a standup, a roadmap, or a kanban of its work.
Each figure on the page comes from one of two places: the crewmate's crew log, or a value the crewmate wrote itself, which carries the label crewmate wrote this. A cost nobody measured shows a dash rather than 0.
The same preview adds a status row above the composer and an All Dashboards page. Its Automatic cards for all sessions switch keeps a summary of each session's recent work and next steps.
The composer warns when a prompt nears or passes the model's context window. A send over the limit is held until you send it a second time, so you can trim the prompt before it goes out.
Turn on Settings → Chat → Composer → Style Markdown While Typing to see bold, italic, strikethrough, and inline code drawn as you type. It is off by default because it switches to a rich-text input that is still in beta. The plain text box also continues a list on the next line.
Three more ways in: drag a file or folder row from the Files tab onto the composer to mention it where you drop it. With the composer empty, press Cmd+Up (macOS) or Ctrl+Up to edit your last message. Type ! <command> to review and run a shell command in the session's workspace.
Hover a message and choose Quote, or right-click it, to carry the whole message into your next send as a quote card that jumps back to the original.
A table in a reply gets a Widen toggle that spreads it into the pane's spare width. Its columns drag wider or narrower from the edge of each header cell.
A settled Side Chat answer offers Add to main chat, which puts the answer in the main composer for you to send. A hung side turn can now be stopped.
The filter menu hides Unread, In progress, or Pinned sessions with an eye-slash button, and Pause all filters lifts every status chip at once. A pinned session stays in view while a tag or status filter narrows the rest.
The sidebar header's menu offers Clean up empty folders, which shows the folders it will remove before you confirm. Clicking an open folder's connector line collapses it, and a folder path written in chat becomes a chip that reveals the folder.
A session's row menu now offers Regenerate title. Turn on Settings → Notifications → Desktop alerts → Mark sessions unread only when they need you to mark a background session unread only when its agent finishes or waits on you.
Turn on Settings → Developer → Feature Previews → Remote crew sessions and a peer crew's session opens as a window onto that crew's own transcript, running state, and approvals. Nothing is stored on this machine. The Sessions list puts Local first, then one collapsible group per crew with an Online, Reconnecting, Error, or Offline badge, so you can see which machines are reachable before you open a session.
Turn on Settings → Messaging Channels → Slack → Connect new sessions to Slack automatically and a persistent session you start in the dashboard gets a thread in your direct message with the bot after its first ordinary message, when Slack is connected. A session whose first input runs as a slash command is not linked automatically, and channel, remote, incognito, and temporary sessions are not linked.
Telegram serves each forum topic in a private chat as its own session, and TELEGRAM_API_BASE_URL routes the Bot API through a reverse proxy where the public host is blocked.
A Telegram or Teams message that reaches a session mid-turn is steered into the turn or queued, and the conversation tells you which.
The Usage tab gains an Estimated Tokens card and an Est. tokens column in Daily History. Kiro CLI reports credits but no token counts, so these are estimates. When enough usage data is available, the Kiro Account modal says under its meter whether your credits last until the reset or, at your current pace, the day they run out. When the provider reports credits, a subagent run shows them beside its elapsed time, and a cron run's result carries the same usage footer as a chat turn.