Automatic page contents, branded agent conversations, and a smarter Operations review
Give your pages automatic navigation menus, have real conversations with your agents, and review proposed record changes in a cleaner, more reliable interface.
What's new in Hub
This update spans three areas: a brand-new automatic contents feature for site pages, a major upgrade to how you interact with your agents, and a much cleaner experience for reviewing Operations agent proposals.
Automatic table of contents for any page
Long pages are now much easier to navigate. You can enable automatic heading navigation on any page or record template and choose from three styles:
- Outline — a clean, indented list of headings
- Numbered — headings numbered automatically for step-by-step content
- Reading guide — a progress-oriented sidebar for long articles
Contents can appear in a left or right sidebar, or above your content on narrower layouts. You control which heading levels to include and how deep to go. On mobile, the contents panel collapses neatly so it stays out of the way.
As visitors scroll, the current section is highlighted automatically. And if you enable "Keep visible while scrolling", the contents panel stays pinned to the top of the viewport — even as the rest of the page scrolls beneath it — with heading jumps adjusted so the target section always appears below the sticky bar.
Contents settings travel with your page: they survive duplication, version restoration, and reordering. Each section can have its own position setting, and if a section is deleted the fallback is handled gracefully.
Talk to your agents — in their own voice
Every Operations, Service, Business specialist, and Executive Lead agent now has a conversation interface directly on its page.
You can ask an agent to explain what it does, why it made a particular decision, or how it could be configured better. The agent draws on its own saved settings and run history to give you a grounded, relevant answer — not a generic response.
Agents now also have editable colours and icons alongside their name, so they're easy to tell apart in a workspace with several automations running. Their visual identity appears consistently across the directory, settings panels, and conversation view.
The conversation experience matches Ask Hub: new chat, history, copy, retry, and full-screen controls are all there. Starting a new conversation archives the previous one, and drafts you're working on are preserved if you switch to another surface and come back.
A unified agent directory
The Agents section now opens on a cross-family team overview — showing all your Service, Operations, Business, and Executive agents together in one place, with their purpose, data scope, enabled state, and latest outcome visible at a glance.
Pending reviews and run failures are shown distinctly so you can act on what matters without digging through individual agent pages. Each agent has a compact directory card with consistent layout and a settings panel that keeps the most important configuration — destinations, permissions, approval requirements — immediately visible.
A clearer, more reliable Operations review experience
Reviewing proposed record changes from an Operations agent is now significantly easier.
Proposed changes come first. Instead of wading through a long summary, findings, and source lists to find the actual proposals, the review screen now leads with a change count, the review state, and the approval actions. The summary, findings, and sources are still there — just collapsed into expandable sections so they don't get in the way.
Source groups are easier to read. Run sources are now grouped by where they came from, with five entries shown initially and a Show all control if there are more.
Agent identity is clearer. Each Operations agent now has a dedicated Identity tab — before the Job tab — where you set its name, colour, and icon.
Error messages actually explain what happened. A proposal that couldn't be applied because of a permission issue previously returned only "internal". The error now explains exactly why no changes were applied and what to do next.
Larger collections are now supported. The target limit has been raised to 50 records per run (up from 10), so agents working across a product catalogue with dozens of linked records can gather all the evidence they need. The 500-record scan ceiling and the 10-proposed-changes-per-run limit remain in place.
Evidence is handled more efficiently. When the same record appeared in both the selected targets and the reference set, it was being counted twice — sometimes pushing large jobs over the evidence limit before generation could start. Each record is now counted once, keeping the full content of your selected targets available throughout the run.
More improvements are in progress. We'll keep you updated as they roll out.