What Shipped This Week: Unified Specialist Settings

One settings modal for every specialist

Managing a team of AI specialists used to mean hunting through different screens depending on where you clicked. Open settings from the Agents app and you got one experience. Open it from chat and you got another. Hire a new specialist and you landed somewhere else entirely.

We collapsed all of that into a single settings modal at the shell level. Whether you tap a specialist card on the Team board, click the title in an active chat, or finish hiring someone new, you get the same panel: Overview, Duties, Persona, and Apps tabs — with the full duty editor built in.

The duty editor is the part we cared about most. Schedule presets (daily, weekly, custom cron), publish mode, chaining to the next job, webhook URLs you can copy in one click, and a running poll so you know when a duty is mid-flight. The Apps tab now surfaces skill help from the catalog, warns you when credentials are missing, and lets you set a Google Search Console site URL without digging through menus.

If you have been putting off configuring a Research agent's morning duty or wiring a Social agent to post on schedule, the path is shorter now. Open any specialist, go to Duties, and the whole workflow is in one place. We wrote about why scheduled duties matter in our 24/7 scheduling guide — this week we made that guide easier to follow inside the product.

Duty notifications that stay with the agent

Routine specialist duties were flooding the global notification bell. Every successful morning scan, every completed report — all mixed in with errors, replies, and system messages. Useful if you want a firehose. Annoying if you run five specialists with daily duties.

We split notifications into two tiers. The global bell still handles errors, support replies, and non-specialist cron jobs. Routine duty completions now show up as per-agent activity: a badge on the sidebar row, a count on the Team board column, and a pin-style strip above the specialist chat when you have unseen updates.

Tap the strip or the "N new duty updates" link and the drawer opens filtered to that agent. Counts clear when you open the drawer — not when you merely glance at the chat. Your Research agent's Tuesday report does not disappear into a wall of noise anymore.

Also shipped

• PWA push notifications. Install CloudAxis to your home screen and subscribe to push alerts. Duty completions and inbox items can reach your phone even when the tab is closed. Host-aware URLs mean the installed PWA opens the OS inbox, not a raw dashboard page. See our home screen install guide.
• Live duty preview on home. When an agent is mid-duty, a compact glass card appears on the right side of the home screen showing scratch-pad progress. Click through to the chat if you want details.
• Home command bar → Cloudia. Type a prompt on the home screen and it routes to your Cloudia thread instead of getting lost.
• Display name greeting. Set your name in Settings and the home welcome uses it. Persists across refresh and passes through to agents.
• Team board follow-ups. One-shot tasks from Cloudia chat now appear in a Follow-ups column. Every card has Cancel/Delete. No more orphaned duties with no way to stop them.
• Browser and file PiP. macOS-style traffic-light controls on browser picture-in-picture. Global PiP across the desktop shell. File PiP with the same chrome when you minimize a document.
• Signup hardening. hCaptcha on CloudAxis signup, Tor blocking, tighter rate limits, and founder alerts only after email verification.
• WhatsApp notify skill. Agents can send proactive plain-text WhatsApp messages from chat — useful for duty summaries that land in your pocket. Related: agents that text you back.
• Onboarding and login fixes. Tour replay layering, login persistence across visits, and specialist card opens from chat title on the Overview tab.

What's next

Week 5 on the content calendar is a CloudAxis vs CrewAI comparison — but on the product side we are focused on making mobile duty review faster and tightening the Specialist → Duty → Output loop so you spend less time in settings and more time reading what your agents produced.