
Azure, Loops, and an Inbox That Knows What Is On Fire
Most of this release is about making the platform act without being asked — and making what it does legible.
Loops close a gap the automation model has promised since it shipped. Recurring schedules could create work on a timer, but nothing could say "when a critical finding is raised, open a remediation task and draft the ticket." Now a Loop binds a trigger and a filter to plain-language instructions, and when it fires the agent plans and executes the work through the same pipeline every other agent run uses. You see the result in Runs. Loops are edited as drafts and published deliberately, because something that acts on your workspace by itself should not change behavior while you are halfway through a thought.
Azure joins AWS, Microsoft 365, Google Workspace, and the PaaS connectors. It is a separate integration from Microsoft 365 for a reason that is not cosmetic: Entra app permissions grant nothing in Azure Resource Manager, so an Azure-only customer should never be asked for tenant-wide directory access they do not need. Setup was rebuilt around one observation — the token episki already holds carries the service principal's identity, and the tenant root group's id is the tenant id — so there is nothing to look up and nothing to paste that can fail on a permission unrelated to the task.
The inbox now answers the question an inbox exists for. It used to list everything you owned grouped by kind, so "assessment" outranked "task" whether or not anything was on fire. The Priority tab orders by urgency, shows why each item is there, and the sidebar badge finally reaches zero.
- Loops — event- and schedule-driven agent automation, capability-gated, with draft and publish
- Azure cloud posture with one-click role assignment
- Priority inbox ranked by urgency, with per-member signal choices
- Versioned evidence — one record per stream, re-verified on every collection, no duplicate rows
- Command sees attached images
- Per-channel notification settings and a badge that counts what matters
Cloud Evidence That Actually Evaluates
Multi-account AWS, Supabase, Vercel, Netlify, and GitHub connectors now write real control verdicts — and a failing check raises a finding instead of quietly passing. Plus agent skills, approved exceptions that satisfy a check, per-boundary program reporting, and a desktop app with tabs.
Security Questionnaires, Answered From What You Have Already Said
Inbound security questionnaires land as work items and draft their answers from your trust center and an answer library of everything you have approved before. Plus trust center subscribers and updates, custom control mapping, per-scope tests, evidence-grounded PCI assessment drafting, and findings that read like findings.