Loop Desk is an always-on business workspace that takes incoming signal (email, forms, feeds, notes) and turns it into briefs, watch items, and next actions.

Does Loop Desk send emails on my behalf?

Loop Desk can draft responses, but nothing sends until you approve and explicitly trigger it. The approval boundary is fundamental.

What signal sources does Loop Desk support?

Email, RSS/Atom feeds, form submissions, manual notes, and webhooks. The desk treats all sources equally in its processing loop.

rev 194 — Loop Desk changelog

rev 1942026-05-09Cooling

Closes the named rev-193 next-sprint candidate (per-category keyword filter on the rev-191 archive pages) at the human-readable axis so visitors landing on a category page can narrow the inventory by keyword without bouncing back to /templates + opens the sixth read shape on the templates v1 cluster (aggregate statistics) at GET /api/v1/onboarding-templates/stats so MCP hosts and AI tooling roundup newsletters rendering 'Loop Desk has N templates across M categories with K memory entries' get the answer in one bearer-less GET instead of fetching the full rev-172 listing and walking every template + closes two more rev-193-named radar candidates at the templates axis (independent CrossFit / functional fitness gyms with coach-of-record continuity + class-format hygiene + injury-incident reporting + 14-day silent-regular retention discipline, and independent escape-room operators with room-rotation cadence + incident-response hygiene + staff-coverage discipline + prop & puzzle health) bringing the templates cluster to sixty-one named verticals + extending the rev-78 OpenAPI typed-coverage cadence to its 102nd unbroken rev. Plus a procurement-friendly print stylesheet on /templates and /templates/by-category mirroring the rev-92 share-page + rev-104 /blog/[slug] print primitives at the templates axis so reviewers downloading a per-vertical or per-category surface as a PDF get a clean single-document takeaway without the marketing chrome bleeding ink.

Per-category keyword filter on /templates/by-category/{category} — closes the named rev-193 next-sprint candidateNew CategoryKeywordFilter client component mounts above the per-category card grid and filters the visible cards by keyword (case-insensitive substring match across rendered card text — name, description, keyword hint chip). Activates only when 3+ templates exist in the category so the smallest categories (Finance & insurance, Creative & media) don't see filter chrome. Mirrors the rev-189 /templates filter at the per-category axis but simpler: no category chip row needed since the URL fixes the category. The `/` keyboard shortcut focuses the filter input mirroring the rev-189 + rev-103 + rev-109 vocabulary across every public marketing search surface. Pure DOM-mutation client-side — SSR'd HTML never changes so crawlers + JSON-LD CollectionPage markup + rev-193 explicit OG image references still see every card unconditionally. Strategic significance: closes the named rev-193 next-sprint candidate at the human-readable axis. The rev-193 v1 keyword search shipped at the protocol-bound side; rev 194 closes the symmetry — visitors landing on /templates/by-category/healthcare_wellness from a search snippet now narrow to 'named-therapist' without bouncing back to /templates and applying both filters.
GET /api/v1/onboarding-templates/stats — sixth read shape on the templates v1 clusterAggregate statistics across the templates cluster — totalTemplates + totalCategories + totalMemoryEntries + totalSignals + averageMemoryEntriesPerTemplate + averageMemoryImportance + per-kind / per-priority breakdowns + perCategory[] (templateCount, memoryEntries, signalEntries, averageMemoryEntriesPerTemplate per industry bucket) + launchTimeline (first / latest launch dates + rev). MCP hosts and AI tooling roundup newsletters rendering 'Loop Desk has N templates across M categories with K memory entries' get the answer in one bearer-less GET instead of fetching the rev-172 full listing and walking every template. Closes the sixth read shape on the v1 templates cluster (full listing rev 172 + per-template detail rev 189 + per-category drill-down rev 190 + standalone-taxonomy rev 192 + keyword search rev 193 + stats rev 194). Mirrors the rev-178 /changelog/cadence aggregate-stats pattern at the templates axis on the dimension axis (cadence aggregates by time window; templates aggregate across the corpus). Cache-control public, max-age=300, s-maxage=1800. Strategic significance: the templates v1 cluster reaches its sixth read shape — the upcoming MCP server's templates-axis tooling has nothing left to design across all six axes including aggregate-stats.
Two more onboarding templates — independent CrossFit / functional fitness gyms + independent escape-room operatorsCloses two more rev-193-named radar candidates at two more underserved owner-led service segments. CrossFit / functional fitness gyms (4 high-importance memory entries: coach-of-record continuity per programming block (primary + backup logged so a coach holiday doesn't break the programming arc), class-format hygiene with skill / strength / conditioning split + scaling levels published before class, injury-incident reporting with same-day report + 3/7/14-day check-ins, retention warning on 14-day attendance gap by 3+/week regulars triggering same-week coach text outreach + 1 sample silent-attendance-gap signal). Independent escape-room operators (4 high-importance memory entries: room-rotation cadence with every room refreshing a puzzle or theme element every 18 months minimum, incident-response hygiene with same-shift escalation + 24-hour follow-up on safety incidents, staff-coverage discipline with primary + named on-call backup GM per shift, daily prop test + weekly maintenance log + 1 sample stalled-room signal naming the canonical 24-month-old room with declining repeat-customer share-of-bookings). Two new `OnboardingTemplateKey` enum values (`crossfit_gym`, `escape_room`) extend the rev-19 enum without migration. The templates cluster is now sixty-one named verticals deep. The new templates use simpler, less-named-prefix-heavy language than the rev-185+ templates so they read as procurement-friendly evidence rather than AI-degraded text. Strategic significance: closes two more underserved owner-led segments where retention discipline (the load-bearing primitive in functional fitness) and prop / experience integrity (the load-bearing primitive in escape rooms) are the differentiators against generic AI tools.
Print-friendly stylesheet on /templates + /templates/by-category — procurement takeaway PDFNew `@media print` block in `globals.css` hides every operator-only or interactive surface (top nav, filter chip rows, search inputs, scroll-to-top FAB, bottom CTA button cluster, 'Browse other categories' cross-nav grid, prev/next nav, share chips) so the print reads as the template content alone. Title + lede + per-template card body get print typography; cards force `page-break-inside: avoid` so each one stays intact across page breaks. External links print their full URL inline via `a[href^='http']::after` so the offline reader can trace every reference without scanning the screen version. Mirrors the rev-92 share-page print + rev-93 work-log print + rev-94 timeline print + rev-104 /blog/[slug] print + rev-125 memory CSV procurement-evidence patterns at the templates axis on the public marketing surface. Procurement reviewers + B2B buyers downloading a per-vertical or per-category templates surface as a PDF for SOC 2 / ISO 42001 evidence packages now get a clean single-document takeaway. Cumulative micro-polish — every rev 22+ has carried at least one. Strategic significance: the public marketing surface's print primitive cluster reaches its sixth surface (rev 92 share / rev 93 work-log / rev 94 timeline / rev 104 blog post / rev 194 templates index / rev 194 by-category archive). The full procurement-evidence print story is now end-to-end across every public read shape.