Feature guide
Everything ClickSorcery can do today — organized by topic for owners and evaluators. Shipped features only; no version numbers or release dates.
Jerry and a full agent company — chat, delegation, memory, and model control.
IT Troubleshooter is a chat-only advisor: Equipment shows Brain as required and Telegram/Voice as not needed. Tasks tab lists the problem areas it handles (Jerry, Telegram, OpenClaw, license, vault…). No fake optional skill tags — behavior lives in the agent prompt (one step at a time, redact logs, hypothesize before deep fixes).
Personal coaches panel states where greeting and tone are edited (persona vs SOUL/Rules). Chat errors for LLM request caps explain the safety limit in plain language instead of only the raw multi-tier dump.
Give Fitness Coach a training plan or Nutrition Coach a meal plan and they write Obsidian Markdown you can open. Log workouts (sets, weight, feel) and meals so the next session continues from real numbers. Quick Actions under What can I ask?
Give a coach a nickname in Settings → Personal coaches; the UI shows “Language Coach Melvin”. On Mobile Jerry say Melvin or Language Coach to switch. Full Kwan checklist marks this item shipped.
After Jerry-only Magic Installer, C-suite and specialists that still show a lock can be turned on with the checkbox next to their name. Org Chart + Add agent also lists specialists already drawn under that leader.
Guides and Jerry’s knowledge now match the app: 40 agents, 56 guided tasks, 86 project templates, Chairman you-vs-agent loop, and the new voice layout. Landing and PRODUCT_FACTS use the same generated roster counts. Quick Actions and Product Brain search catch “what is ClickSorcery?”, integrations, and RAG on the Memory tab.
Every catalog project lists the agents, MCP/integrations, programs, and typical cost of each door — plus how it works after you connect. Fifteen new guided templates cover fridge meals, Tailscale, vault RAG, API-to-code, home energy (DIY), and Mobile Jerry P2P. You still open accounts; agents do not buy them.
Short questions and live coaches use the fast Groq/voice stack; OpenClaw streams words into the chat; the website explains what you can speed up and what a plan will always take longer.
Short questions and live coaches answer on the fast cloud/Ollama stack; OpenClaw text streams into the chat; a readiness bar shows gateway warm-up and restart.
Invoicing & Billing agent: pdf-parse + local model extraction, Approvals before save to Finance/Invoices in Obsidian, copyable accounting drafts — no cloud accounting API.
Installer release for four new specialists (BDR, voice receptionist, invoicing, tax monitor), 40-agent roster sync, Jerry orchestration knowledge, and Phase B roadmap backlog.
Sales BDR, Voice Receptionist, Invoicing & Billing, and Tax Compliance Monitor join the roster under CMO/CFO/COO — drafts and checklists first; live telephony and accounting APIs stay Owner DIY until wired.
Each agent Dashboard and Setup → Activity show one filtered timeline: config changes, logs, LLM cost, delegations, approvals, and mission steps.
Setup → Identity & trust: hierarchy, persona, circle of trust, and rules. Per-agent HITL levels now shape chat turn context and mission external acts (approve vs auto vs notify).
Each agent Setup tab now groups base vs optional skills, integrations with install links, tasks & projects, comms/vision honesty, and per-agent HITL on Costs — with (i) info modals.
When Jerry’s Home checklist is complete, Org chart unlocks + Add agent: pick a C-suite lead, activate a catalog specialist or create a custom agent, then follow the same brain/voice/budget checklist.
Each agent now has four main tabs — Dashboard, Setup, Memory, Costs — plus a Projects link. Setup holds equipment, editable tasks, skills, instructions, and advanced config.
When Jerry uses Hybrid + local Ollama, a chat bar switches the next replies to cloud Auto multi-tier until you return to local.
Jerry brain setup starts with where AI runs: local Ollama, cloud (Auto or one direct provider), or hybrid. Cloud path is two cards, not six providers at once; optional monthly cloud cap slider on the same screen.
Magic Installer skips the agent roster — Jerry and Layer 0 start active; specialists stay locked until you add them. Home shows a setup checklist; low-RAM PCs auto-pick cloud and hide Ollama in Task 1.
Jerry project plans use a two-stage pipeline: compact consultant digests, then a schema-backed plan with targeted repair instead of blind regeneration. Zapier, HubSpot, and similar externals show up in preflight as Owner setup — not as invented ClickSorcery features.
Seven personal coaches get 2026 system prompts and model routing; Speak Up Coach joins the roster; per-agent TTS/STT persona presets live in Settings.
Model Agent recommendations now follow the agent catalog (CFO/research pinned, routine agents Auto multi-tier). Dynamic task routing is on by default so hard prompts escalate to stronger tiers.
Fitness, nutrition, time, debate, language, and philosophy agents share Obsidian templates so sessions continue from real numbers and chapter progress — not chat amnesia.
Jerry’s pack card now activates for real and continues your request. Office Basic and GIS/QGIS are available as prompt packs; OpenClaw Operator stays locked.
Jerry is the primary chat partner; voice mode is the default — large mic, push-to-talk sends immediately, TTS reads replies aloud.
When Jerry consults a specialist mid-chat, that agent runs in its own OpenClaw session — memory stays scoped, not mixed with Jerry's thread.
Long tasks run in the background — Jerry notifies you in chat when results are ready; Mission Control shows the queue.
Choose Ollama, pin a cloud provider, or confirm Auto multi-tier routing — Jerry chat stays blocked until brain path is chosen.
Your main chat partner coordinates the agent company: answers questions, routes work, and keeps context across the team.
C-suite (COO, CFO, CMO, CIO, CTO), Security Layer 0, Model Agent, and specialists — all visible from day one with org chart.
Every agent has the same deep profile UI — instructions, memory, skills, configuration, runs, budget, tasks, and equipment.
Visual company structure with drag-and-drop layout — see who reports to Jerry and how teams connect.
Consult CFO, CIO, or another agent for advice without leaving Jerry chat — real answers from that agent's context.
Jerry can draft a project plan for approval, then fan out resource requests after you approve — with LLM intent detection in EN/ET phrasing.
Store who you are, how you communicate, goals, and knowledge documents — injected into every agent prompt automatically.
Review and approve model assignments per agent — batch or single — before changes apply to the roster.
Live OpenRouter pricing for nine curated models (three per routing category), with cheaper-or-faster recommendations per agent.
Background checks on Vault health, skill/tool review, and account isolation — runs without blocking your daily chat.
Control when long chats are summarized — message threshold (10–500), and modes ask, auto, or never.
Discoverable dropdown in Jerry chat for model review, project consult, routines, Telegram send, Spotify, agent consult, and more.
Whisper STT and TTS in chat — text-only, voice-only (push-to-talk auto-send), or both; per-session read-aloud toggle.
Index your Obsidian vault markdown and inject relevant excerpts into Jerry chat, fallback chat, and Telegram replies.
When an agent uses Brain = Auto, ClickSorcery tries Groq → DeepInfra → OpenRouter by role — with failover on 429/500/503.
Bring your own search API keys — agents can look up live information when keys are connected in Vault.
65 integration names/descriptions and 33 agent profile strings localized — Estonian owners see native labels across catalog and detail panels.
Connect accounts and keys in Vault — live detail panels when credentials are set.
Phone companion gains stop-speaking and mute-reply-voice controls. Proactive day “cards” are labeled as screen alerts; mic settings as connected devices.
Phone opens straight to chat with a text bar, hold-to-talk mic, and hands-free toggle. Photo/voice answers follow desktop Voice language (Estonian text and speech when Voice is ET), not English captions spoken with an Estonian accent.
Mobile Jerry: Type / Hold / Live after pairing; spoken text hits phone chat and the home desktop as soon as STT finishes. Free Edge neural TTS works again (Sec-MS-GEC v2). Voice settings: six languages with free Edge default; paid preset no longer changes listening.
Team agents turn on from the header checkbox without finishing the Home checklist or joining a project. Integrations and Vault now list Merit Aktiva, Xero, QuickBooks Online, and Sage — credentials only; Jerry does not post invoices to those ledgers yet.
The phone camera preview stays on the chat screen so 📷 can take a real frame. Car Bluetooth is audio only — it does not turn the camera on or off.
Mobile PWA: chat-first Kanal, bottom tabs, settings for agent/mic/camera; invoice PDF parse + Approvals; Pro license key spacing fix for API Vault.
Phone companion: one chat screen, bottom navigation, settings for agent/mic/camera/proactive day, pending-approvals badge on Kinnitused.
Pair your phone with a QR code; reach the home desktop hub over Tailscale or your HTTPS tunnel — not only same Wi‑Fi.
Talk to Jerry from your phone through the home desktop hub — pairing code, push-to-talk, STT and TTS. Same Wi‑Fi still works; Tailscale or a tunnel is the away-from-home path.
Install Jerry on your phone home screen (PWA) and see live status over WebSocket while a voice turn runs.
Build Android/iOS wrapper from mobile-native/ — full-screen mic app with deep link pairing from desktop.
Unified sidebar with search, connected/needs-setup filters, collapsible categories, and structured detail pages with live Vault panels.
Browse every integration with honest status — connected when Vault is set, or setup guide when not yet wired.
Connect a Personal Access Token — see repos and open issues; Jerry can answer questions about your codebase context.
OAuth connect — today's events and the next few days show in the integration panel when linked.
OAuth connect — unread count and recent unread messages (metadata) in the detail panel.
OAuth connect — storage usage and recently edited files without leaving ClickSorcery.
Read-only revenue view — MRR, 30-day revenue, balance, and recent charges for finance agents and Morning Brief.
Buy ClickSorcery for €79 on clicksorcery.com — Stripe Checkout, webhook, Supabase purchase record, and license key email on success.
Today's tasks, add new tasks, and mark complete — plus Morning Brief lines for priorities and overdue items.
Search, read, and create pages from your workspace when an integration token is in Vault.
Browse bases, tables, and records; create records where permissions allow.
Live weather for your city — free-tier API key, Estonian descriptions when preferred, Morning Brief weather line.
Verify bot or workspace tokens and list channels — read-only today; posting flows stay approval-gated.
Profiles, balances, and recent transfers for finance visibility when API token is connected.
Project list and deployment state — failed deploys highlighted for quick ops checks.
Username, recent projects, and open merge requests when a personal access token is in Vault.
Project overview and table inventory via service role key — useful for dev and data agents.
OAuth connect — now playing and playlists in the panel; ask Jerry to search and play tracks (Premium) in chat.
Channel stats and recent videos when linked through Google OAuth.
Profile metrics and recent posts with a Bearer token — read-only social snapshot.
Name and network metrics from the API — read-only professional snapshot.
Follower count, media count, and recent posts for connected business accounts.
Phone metadata and quality rating from the Business API — not full inbound chat in MVP.
List workflows on your instance, search n8n.io templates, and trigger runs through Workflow Integrator after approval.
Send and receive text and voice notes via bot token — Jerry transcribes voice, replies with text or TTS audio, with processing ack while thinking.
Routines, Morning Brief, and Workflow Integrator for n8n.
All catalog templates follow the same loop: Start with Jerry, agents prepare, you open accounts and keys, then Stop when you do not need the project.
Tell Jerry to automate something — he matches an n8n workflow, suggests templates, asks for confirmation, then triggers only after you approve.
Built-in schedules for Morning Briefing, End of Day, and Weekly Review — or propose new routines through Approvals.
Daily briefing routine that pulls live blocks when integrations are connected — set the time and let Jerry summarize your day.
Vault, security, brain, channels, and personalization.
Free Edge neural TTS uses Microsoft’s Sec-MS-GEC v2 token so Play sample works (Russian Svetlana, Estonian Anu, and other locales). Settings flow is language first, then Text / Free / Paid; switching to paid no longer changes listening (STT).
Settings → Voice picks agent/speech language (ET, EN, DE, RU, ES, FR). Free Microsoft Edge neural voices apply per locale; OpenAI/ElevenLabs/Cartesia are optional in advanced. Jerry replies in the voice language even when app menus stay Estonian.
Equipment → Voice is language-first with listen/speak cards, mic level bar, and Play sample. Jerry and Help guides match 40 agents, 56 tasks, and 86 projects. Settings → License copy works via desktop clipboard. Download roster files updated to 40 agents.
Equipment → Voice starts with the agent language. Presets stay as a compact row. Listening (STT) and Speaking (TTS) sit side by side with a key badge, a voice list filtered for that language, and Play sample. The mic test and a level bar are right under the cards — not buried in extras. Obsidian RAG stays on Memory.
Settings → License → Copy key writes the key to the clipboard via the desktop process. The old browser clipboard call was blocked in the sandboxed window, so the button appeared to do nothing.
Talk with Jerry without a paid TTS key: Edge-TTS (Microsoft neural, Estonian Anu/Kert, internet) or Piper (local English after a one-time download). Cartesia/ElevenLabs/OpenAI stay optional. F5/XTTS GPU cloners are not in the app.
Save a Cartesia key in Vault and pick Cartesia under Jerry → Voice. Replies stay on Cartesia. A library voice is chosen if you typed a name instead of a Cartesia voice ID.
If OpenClaw is not on a new PC, chat no longer pretends it is a bad API key. ClickSorcery installs the engine on first launch (Node.js + npm), or from Install OpenClaw on the readiness bar / System health. You can still chat via Ollama or a cloud key while that finishes.
Jerry chat push-to-talk: hold large mic to record; Whisper path no longer conflicts with broken browser STT in Electron on Windows.
Settings → Voice info modal: English copy no longer implies Estonian-only setup when describing free OS text-to-speech.
Settings → Voice explains listen → think → speak, costs, and why free OS TTS is weak for Estonian; recommends Groq STT + cloud TTS with local Ollama.
Public onboarding page after install; agent skill (i) prompts with examples; read-only model tier hints on Brain config.
After license validation, the app uses Stripe purchase country for timezone (EE → Tallinn, US → Eastern, etc.), with manual override in Settings → Profile.
Settings → Safety shows gateway status, opens the gateway log, restarts with brain sync, and lists copyable openclaw doctor / gateway CLI commands.
Brain setup adds path D: one router hub and a fixed model, separate from Auto multi-tier failover.
New Costs → Chairman overview: estimated monthly total by provider, manual Cursor/ElevenLabs notes, per-agent brain and voice spend, agent ON/OFF, and pause proactive coach.
Install v1.0.21 for reliable /download installers, clearer errors when only Groq is in Vault, and guided local Ollama brain setup in Magic Installer and Brain.
Install v1.0.20 for app-wide monthly cloud spend limits (€5 default on ≤16 GB RAM), FK-E voice coach switching on Mobile Jerry, and RAM-aware warnings on heavy tasks/tools/projects.
Set an app-wide EUR cap so paid cloud (OpenRouter, Claude, etc.) cannot exceed your limit. Free Groq stays ~€0. New desktop installs on ≤16 GB RAM default to €5/month — change anytime; 0 = unlimited.
Unified catalog routing for planning, synthesis, and critic. When a provider rate-limits, plan-quality steps wait briefly on the same model, then fail over to the same quality class on another provider — not to a weaker 70B or local fallback.
“Talk now — free” always uses browser STT + system TTS; System selection no longer silently switches to Cartesia. Cloud preset copy clarifies OpenAI API usage.
Agent Equipment and guided tasks use the same card pattern: what / when / cost / how — with Advanced accordions for power users.
When Obsidian is installed and the vault is live, Settings shows a linked status instead of a perpetual “Connect Obsidian” shortcut.
Passphrase-encrypted backup restores Vault keys and agent memory after a disk failure. Marketing agents keep brand/campaign files in the vault. Skill pack prompts can be edited in the app.
Speech-to-text can use Groq Whisper Large V3 Turbo with the existing Groq Vault key. Live TTS prefers Cartesia when key and voice ID exist; system voice stays the honest install default.
Magic Installer no longer asks for Jerry's LLM twice; system agents (Jerry, Model Agent, Security) start active on team preset.
When OpenSSL is available, the mobile hub serves HTTPS and WSS with a self-signed certificate on your local network.
Add family profiles on one desktop — members talk to Jerry with voice; owner keeps Vault, Approvals, and admin settings.
Long Jerry answers play sentence-by-sentence — the next segment synthesizes while you listen (OpenAI, ElevenLabs, Cartesia).
Documented Whisper STT and TTS setup (OpenAI, ElevenLabs) with Jerry knowledge and Quick Actions.
Live CPU, RAM, and disk in the status bar — plus today/month API spend by agent and model. Click the System or Budget chip for the full panel.
All keys encrypted in the main process — masked in UI, never in localStorage or plain .env files.
Allowed and blocked actions, human-in-the-loop levels, and onboarding security briefing before first launch.
Workflow Integrator defaults — confirm before every run and maintain a trusted workflow ID list.
Telegram bot token, Jerry email field, voice provider status, and Brain setup wizard entry point.
Set vault path, see file counts, open folder, export ZIP backup, and restore — central memory for agents.
Name, city (for weather), nationality, hobbies — feeds Jerry's Brain and Morning Brief context.
7-day trial and upgrade path in Settings — activate your ClickSorcery license without leaving the app.
First-run Terms modal with €79 pricing and agent-permission summary — links to live Terms and Privacy on clicksorcery.com.
Updated Terms §5 (agent defaults, permissions) and Privacy §10 (DPF/SCC) — single source on /terms and /privacy, linked from download checkbox and desktop ToS modal.
Projects, approvals, memory files, onboarding, and ops.
Desktop Pro stays €79 as early-bird / pre-sale for November–December 2026. The pricing section now lists Full Kwan work ahead (nicknames, training diaries, calendars, hands-free, and more) with checkmarks as items ship — Full Kwan price is not announced yet.
Click 40 agents or 86 projects on the homepage. /agents lists Jerry plus every specialist. /projects lists what you can build with a guide — and what you do vs what the agent does after you connect.
Web /first-steps reframed: Home, Brain, Voice, and mobile pillars first; Jerry interview moved to step 10; Voice FAQ link.
Interactive glossary adds BYOK, Model Router, Vault Steward, voice (STT/TTS), Tailscale, Approvals, full stack terms, hardware tiers, 56 tasks vs 71 projects, and 2026 home UX — EN and ET.
Redesigned home: hero, setup stepper, live activity rail, solo Continue setup CTA while onboarding. Playwright home grid E2E included.
New /first-steps page: Home, Brain, Voice, and mobile connection first; license and Magic Installer; Jerry interview as step 10; links to setup, brain, and voice FAQ.
New homepage section and FAQ set expectations: guided onboarding railway vs executive project menu, Chairman Approvals, no grind-to-unlock — honest about integrations and n8n.
Redesigned home: hero, setup stepper, and a live activity feed on the right (refreshes every 5s). During onboarding, one clear Continue setup button — no stat clutter or extra shortcuts.
New /local-models page: official Ollama only, model picks, local vs cloud table, setup checklist, RAM tiers. Site footer uses three link columns.
Rename Jerry to your operator name (e.g. AADU) and UI copy follows — dashboard, tasks, voice hints, heartbeat, settings (EN + ET).
Your chosen CEO name (not Jerry) shows in chat banners and heartbeat; taller chat until voice is set up; top-bar update strip when a new version downloads.
Skill, integration, and task (i) popups with prompt examples; Brain model hint panel; install from /download after mirror promote.
Geo-based €79 / £79 / $79 Stripe checkout; agent Integrations and Tasks tabs get (i) prompt examples like Skills.
Jerry no longer hijacks normal chat with Vault Steward when Ollama is off; MVP banner and landing honesty FAQ for routines and vault.
Hero MVP circle opens a short Minimum Viable Product popup; header languages use a dropdown like the desktop app.
Mission Control shows live work first; automatic timezone from Stripe purchase country; clearer Vault Steward errors when Ollama is off.
Install 1.0.26 for System health OpenClaw troubleshooting, single-router brain setup (Groq/DeepInfra/OpenRouter only), and gateway/brain sync fixes.
Dream Home and Journey now recommend Mac mini M5 Pro 64 GB with 10GbE for department nodes; Boss stays Mac Studio Ultra 192 GB+ (verify on Apple).
Header language codes (EN–RU) stay on one row on desktop; narrow phones use the mobile menu row.
Install 1.0.25 if Agents → any profile crashed the UI (React hooks error). Activity timeline and agent card from 1.0.23+ remain included.
Landing, FAQ, and /download roster files now match the desktop catalog: 36 agent profiles and 56 guided tasks — synced automatically from the app registry.
Install 1.0.23 for the Pro agent setup experience: grouped skills/tools/comms, Identity & trust tab, per-agent human-in-the-loop in chat and missions, and a unified activity timeline on each agent.
Install 1.0.22 for Obsidian Vault Steward (approve-before-write), add-agent wizard, agent profile tabs, chairman cost overview, and mission triggers on Routines.
Layer 0 Vault Steward (local Ollama) turns pasted text or PDF/DOCX into a structured markdown proposal. You approve in Approvals before anything is written to PARA-lite folders; RAG can index the new note after save.
Header language picker lists all six product locales. Full Estonian site copy where available; German, French, Spanish, and Russian use English page text until localized.
Windows and Mac installers download from clicksorcery.com (not private GitHub URLs that showed a 404 / error page when you were not logged into GitHub).
New FAQ clarifies Mobile Jerry + Fitness Coach for hands-free workouts: home desktop hub, push-to-talk turns, BYOK STT/TTS — not a phone-only realtime streaming coach.
Landing and journey copy: sequential agent use (not 33 parallel jobs), Groq/free ~€0–15 vs premium models, built-in monthly cap, Cloud Commander as a light disk footprint.
Mobile Jerry gets opt-in proactive nudges (meals, supplements, focus), Google Calendar T-30/T-5 meeting reminders, voice agent handoff between coaches, and cloud edge push (FCM HTTP v1) when the native app is away from the desktop hub.
Cloud edge device register, job poll, magic pairing sessions, and desktop Native API v1 scaffolding — secrets stay in Vault on the desktop.
Home page and Google title focus on one AI for work, home, and everyday life — full agent team when you are ready, not “run a company” as the first impression.
Automated honesty checks for workspace save, agent chat, skill packs, projects, and approvals — without calling live LLM APIs in CI.
Approve and Reject update immediately; workspace state refreshes so the queue and shell badge stay honest.
Landing Dream Home and Journey no longer cite M2 Ultra or an 8×4090 NVLink farm; PC path is RTX PRO 6000 Boss + 5× RTX 5090 departments.
After Jerry's brain (Task #1), guided tasks walk you through voice setup then Mobile Jerry — the path to talking without typing, including on the go.
Twenty paste-first ProjectFlow templates are core-verified — fill fields in Projects, Jerry runs the workflow without n8n.
Integration catalog descriptions in Estonian (44 profiles) and localized ProjectFlow form labels for all 20 core-verified projects.
PLAANI PUNKT 5 wave 1 — when UI language is English, all 71 project templates show English names, descriptions, what-you-build text, and steps.
Full PLAANI PUNKT 4 catalog review: 20 core-verified ProjectFlow projects, 50 experimental, status matrix, and auto-generated inventory.
PLAANI PUNKT 6 MVP closure — all 65 tools have Estonian catalog names, task level labels are i18n-aware, and Quick Actions / Settings Help guides expanded.
Expanded guides for Knowledge tab, Company org chart, and Routines flow — plus Estonian display names for 18 key tools in the catalog.
Expanded guides for approval flows, Meeting Board (Ollama-only replies), logs table view, and Security & Costs — plus Estonian agent catalog strings in the UI.
Tools & Integrations and Skill Packs guides updated — catalog health dots vs manual status, and pack activation injects into agent prompts (not just a flag).
Product guides now match the catalog: 71 project templates (20 core-verified with ProjectFlow), 55 guided tasks with auto vs self-confirmed badges.
Product guides and Jerry navigation updated to match the app — voice setup lives under Jerry Equipment → Voice, not a separate Settings panel.
Full agent list with first tasks and guided tasks — Markdown (ET/EN) and CSV from the download page or repo.
71 project templates (20 core verified workflows + 51 templates incl. Jarvis) — ProjectFlow forms on verified core; filter in Projects tab.
Per-agent daily memory notes and LONG-TERM.md in the vault — visible on each agent's Memory tab.
Review project plans, model assignments, routines, Telegram sends, and workflow triggers before anything goes live.
First-run flow: pick roster, security briefing, Jerry's Brain intro, model setup, then wake your AI company.
Activity logs and cost tracking surfaces — see what ran and where API spend goes.
Agents execute through OpenClaw gateway — cloud models by default, local Ollama when configured.
Public site language switcher, Changelog feature guide, and cleaner header — Dream Home, Journey, and Glossary moved to footer only.
Honest Cloud Commander (~1–3 GB) vs Local Supercomputer (~50–100+ GB) scenarios, component size table, and local model download warning — EN/ET.
Public site uses Desktop Pro (€79 shipped product) instead of MVP framing; FAQ and schema.org synced with productFacts (33 agents, disk space, Full Kwan upgrade path).
Brain guide and setup guide use Desktop Pro + catalog counts from productFacts; disk space cross-links; Terms/Privacy tier wording; ET glossary Full Kwan fix (one-time, not monthly).
Landing copy audit complete (Desktop Pro framing); privacy notes Vercel Web Analytics on marketing site; domain smoke test verified key pages live.
Tag a semver release (v1.0.0) to build Windows NSIS and macOS DMG in GitHub Actions — combined draft release with changelog.
Catalog entries with setup guides only — not listed above until live in the desktop app.