Session states now look the same on every dashboard: green for healthy, cyan for working, amber when an agent needs you. Working agents stand out in the 3D aquarium, name tags no longer cover creatures behind them, and Claude Code's arms are repaired. OpenClaw setup guidance appears only when pairing or sign-in actually fails.

Explore an optional 3D aquarium on iOS 18 and later. Familiar agent characters, reactive fish and a wandering snail bring your sessions to life. Tap empty water to hide the panels and move closer; tap again to return. Selected sessions and input requests stay prioritized in crowded scenes, with the full session list available. Your existing dashboard choice is preserved.

• Shows z.ai GLM Coding Plan usage received from your paired Mac, with distinct credit and MCP tool-call readings. • Improves live device previews, including provider marks, usage labels, crowded deck layouts, and round-screen placement. • Displays updated usage and session ordering from the paired Mac more consistently.

• Usage rows adapt to available providers and quota windows. • Subscription plan details are easier to read, with improved compact layouts. • Connection and provider selection behavior is more consistent.

A clearer view of your agents Session and timeline updates now identify the agent that produced them more reliably. Missing usage windows stay empty instead of appearing as 0%. With the updated Mac app, OpenClaw plugin approval requests are visible alongside tool approval requests, and session monitoring recovers more reliably when the Mac dashboard service changes. This update also refreshes the App Store screenshots and previews.

See when a session is waiting on work it started A session whose turn has ended reads "idle" — even while the workers it launched and the background jobs it started are still running. Session rows now mark that: beside the subagent count, a waiting count shows the sessions and jobs it is still waiting on, so a quiet row is not mistaken for a finished one. Attention only when someone is really waiting for you Sessions running in accept-edits or default mode raised a permission alert for tool calls their own terminal never asked about, then stalled for 25 seconds before continuing on their own. The permission-rule matcher now follows the documented syntax, including compound commands and the built-in read-only set, and resolves anything uncertain toward not alerting. Claude's internal helper queries are no longer counted as finished subagents. The usage panel explains a missing quota When Claude credentials expire, the subscription gauges name the cause instead of going blank, and clear it automatically once a reading succeeds.

Pairing with your Mac is more forgiving in 1.2.0: • When the Mac approves this device in its new Pair a Device window, the app reconnects on its own — a refused connection is now a short holdoff instead of a permanent ban that needed an app restart. • The app introduces itself with a stable per-install identity, so the Mac can approve this exact device rather than the network address it happens to hold. • Reliability fixes shared with the 1.2.0 Mac app, including correct badges for Desktop Codex sessions and consistent network discovery identity.

Fixed Codex usage not showing correctly. • The gauge could show one model's quota (GPT-5.3-Codex-Spark) instead of your account's, reading 0%. • Changing your ChatGPT plan no longer makes the Codex gauges disappear. • ChatGPT Pro Lite is now shown by name. • Usage no longer clears when you change a setting.

Kiro joins your deck. Kiro CLI and Kiro IDE sessions now appear alongside Claude Code, Codex, OpenCode and the rest — in the session list, on the timeline, and as their own creature in the terrarium. Also in this release: • A live count of the subagents working under each session • A badge showing which provider actually answered, so a Claude session pointed elsewhere reads correctly • A clearer approval prompt for OpenClaw, readable and answerable from the device

Feedback-flow and stability improvements. • After meaningful use, share feedback directly through Apple's native rating prompt • Added an App Store rating and review link in Settings • Usage-day eligibility stays on device; no session or project information is collected or sent • Stability improvements Thanks for the feedback — keep it coming: admin@foundby.kr

Connection and pairing reliability update. • Clear “Unauthorized” guidance replaces generic connection errors when a Mac requires pairing • A paired Mac now stays paired when its dashboard service restarts or hands over • Local-network discovery never exposes pairing credentials • Stability improvements Thanks for the feedback — keep it coming: admin@foundby.kr

Connection, pairing, and a way to explore without a Mac. • Security hardening: connecting to your Mac now uses QR pairing (discovery still finds your Mac automatically) • Once paired, your Mac stays paired — reconnecting no longer asks you to scan again • Explore without a Mac can now play a demo session, so you can watch AgentDeck work • Device previews scale to fit instead of being cut off • A failed request now shows why it failed instead of spinning forever • Usage labels no longer wrap mid-word Thanks for the feedback — keep it coming: admin@foundby.kr