Back in this tab after an hour: what was it doing, and what did I ask it?
Jump between your prompts. Fold the tool output.
Your last question, without the scrollback hunt.
Your prompts stand out as cards. Jump from one to the next, or use Read one exchange to focus on the current question and latest reply. Fold tool output when it gets in the way. Choose the font, size, spacing and background for each pane. The reading view keeps the agent's words unchanged; it does not summarize them.
Five agents, five half-remembered plans. Which one had the migration?
See conversations across projects, grouped by whose turn it is.
Find the agent you came back for.
The sidebar shows each agent's status. Click 1 needs you · 12m to jump to the one waiting longest. Enable a catch-up line for a conversation to see where it stopped before opening it. The four-conversation reading grid shown here is planned; the current reading view opens one conversation at a time.
You switch to the other one and lose the thread of this one.
Check what the agent said without adding to its conversation.
Ask about the migration without sending another prompt.
Coming next; not in the build yet. Ask a small model about the last few turns and get an answer with quotes linked to the transcript. The coding agent will not receive your question. For now, use the reading view or opt-in catch-up line to find where you left off.
Alt-tab whack-a-mole. And after a relaunch, rebuilding the whole desk from memory.
Drag a divider. Keep the other agents visible.
Give the busy pane more room.
Keep your agents in one window and resize their panes as work shifts. In the example, parser-crash needs a decision while three other panes stay visible. Your layout and running sessions return after an app relaunch.
The agent that sat idle for 20 minutes on a y/n question nobody saw.
Get a notification when an agent needs an answer.
See how long an agent has been waiting.
When winglet detects a question, it marks the agent needs you, starts a timer and shows the question with its options. Desktop notifications bring it to your attention; phone push is opt-in. Completion notices include check output or say no check ran. Some CLI dialogs can still be missed; see the detection limits.
You step away for coffee and forget which agent was working on what.
Check your agents from your phone on the same network.
Read and reply from the couch.
Pair your phone by scanning a QR code on the desktop. See who needs you, open the question and type a reply while the other agents keep working. The phone view connects to your PC over your LAN, without a relay or account.
Tests, ports, branches, CI: each in yet another window you have to go and look at.
Keep the readouts you check in the same window.
Your test results, beside your agents.
Put a live readout next to your agents using a script in any language. Start with bundled examples for test results, git worktrees or conversation status. Plugins that include code wait for your approval before running. The image shows possible custom panels, not a list of bundled integrations. How to add a plugin.
