For beginners, busy families, and power users — one app on your machine instead of ten separate AI subscriptions.
For work, for home, for everything in between. One AI that learns languages with you, trains with you, handles your inbox, and grows into a full team when you need it.
Start the 7-day trial in the desktop app, then upgrade in Settings → License.
7-day free trial · Full access · No credit card required · Windows now · Mac .dmg soon
What you need
In 2026, a website and a ChatGPT subscription aren't enough. Companies that skip the six layers below are doing manual work at the speed their competitors automate everything.
01
AI workflow orchestration
Connect models, databases, and apps so they talk to each other autonomously — not one tool at a time.
02
Real-time voice & messaging
Customer contact happens in real time — voice and chat — not only over email.
03
Vector DB & knowledge sync
AI must know your contracts, processes, and internal docs — not just the public web.
04
LLM gateway & arbitrage
One expensive model for everything wastes money — smart routing sends each task to the right brain.
05
AI-driven dev environment
Every company is partly a software company now — product work must run at AI speed.
06
Wearables & edge capture
Future edge: the real world becomes structured data without manual data entry.
Full Kwan
The full package: trust and synergy between you and Jerry — your vision plus an agent team that learns your business.
You
“Show me the money!”
You bring the vision — your runtime, API brain, memory, tools, integrations, and context. You set direction and define what winning looks like.
Watch the scene (0:46)Agent
“Help me, help you.”
Jerry takes your context, structures it, learns how you work, and optimizes processes in the background — with approvals before anything goes live.
Watch the scene (1:14)Outcome
🏆 “Full Kwan!”
Machine and human work as one system — value greater than the sum of parts. Desktop today; mobile and always-on when Full Kwan ships.
Watch the scene (0:34)Hardware & cost
4 tiers. One product. Your pace.
Tier 1
(8 GB RAM)
Tier 2
(16 GB RAM / M-Mac)
Tier 3
(8–16 GB VRAM GPU)
Tier 4 — Your goal
(Mac Studio 64GB+ / PC 24GB+ VRAM)
API costs
You don't pay top-tier prices for every agent. ClickSorcery automatically picks the right model for each task.
Jerry (CEO) and the code writer use Claude Sonnet for critical decisions — 80% of routine background work runs on super-cheap models (Groq / Haiku / Llama).
In Settings you choose exactly which agent uses which model. Free models only? Sure. Top speed? That too.
System & Budget → set an app-wide cloud spend limit. New installs on ≤16 GB RAM default to €5/month (change anytime; 0 = unlimited). Free Groq stays ~€0 — paid cloud counts toward your cap.
Before you start
ClickSorcery is powerful — but honest.
01
Jerry guides the setup — what to teach, what tools to add, what your business needs. You bring the knowledge, Jerry brings the structure.
02
Full C-suite and specialist team available day one. Each agent has their own brain — the right model for the right job.
See who they are →03
Obsidian memory means Jerry never forgets your context — across sessions, across projects.
04
You control how much Jerry does on his own. Start slow — trust builds over time.
05
Not a demo kit — a real company stack. See what tasks vs projects mean ↓
Browse guided projects →Tasks & projects
Guided onboarding railway + executive menu. Finish what you need — no grind-to-unlock game.
The guided onboarding railway
A 14-level hands-on path: local vs cloud brain, Vault, voice, Approvals, Telegram, agents, and workflows — configured on your real desktop as you go.
You become an AI power user with a tailored digital HQ. Completing every task is optional mastery, not a secret product unlock.
The executive menu
A buffet, not 71 chores. Pick the template that matches today's need — launch plan, marketing calendar, ebook structure, finance outline, automation design.
Press Start with Jerry → concrete plan, structure, and drafts. Use one template today; ignore the rest until you need them.
Realistic execution
Agents handle the heavy lifting — analysis, long-form drafts, checklists, and code sketches. Integrations, n8n, and accounts are opened by you when a template requires them.
Review, tweak, and Approve before anything goes live. No badge turns into 71 autonomous SaaS products running without you.
20 core-verified templates are the fastest start; the rest vary from chat-only to integration- or n8n-dependent (honest doors in-app).
How you actually use it
Want something → open a project → follow the guide → install, buy a key, or create an account if needed → connect → that agent can keep working with it. Connecting is not “the agent now lives your life.”
How it works
ClickSorcery runs 40 AI agents on your desktop — Windows or macOS. Jerry (AI CEO) leads the team. You talk to Jerry in plain language. He assigns work to the right specialists automatically.
What's included from day one:
Pricing: €79 one-time Desktop Pro. No monthly subscription. 7-day free trial. No credit card required.
Fit check
Why a team
Same principle as a real company — specialization wins.
One generalist agent spreads thin. A marketing agent knows your brand voice. A sales agent knows objections. A coding agent does not mix business logic with syntax. Each gets the context and tools built for their job.
Marketing drafts lead magnets. Sales replies to inbound leads. Dev ships the feature. At the same time. No waiting in a single queue.
Analytics spots a trend → Marketing drafts a campaign → Sales says "that does not fit our buyer" → Quality review before anything ships. Agents review each other.
Sales process changed? Tune only the sales agent. Need legal review? Add one node and one model. The rest of the stack stays untouched.
A company with one person doing accounting, marketing, engineering, and sales at once can do a little of everything — but loses to a specialized team on anything serious.
ClickSorcery agents work like team members — each with a role, tools, and ownership.
Your journey
Jerry guides every move.
Jerry starts by learning how you work — your goals, constraints, and what winning looks like.
He maps your context first, then builds a plan around what actually matters.
That's the deal: you give him clarity, he gives you a team that executes — with your approval on anything live.
00
~10 min · free
Magic Installer walks you through Cloud or Local mode.
00
~5 min guided tour
Projects, agents, tools, approvals — one workspace.
00
20 core verified + 66 templates · est. €0.10–2 API cost
Pick a template. Jerry runs the plan with you.
00
Telegram ready · more channels as guided projects
Connect Telegram today. Email, WhatsApp, and others follow the same pattern as you need them.
Integrations
Talk once Jerry's set up. Getting there takes a bit of teaching.
You
Jerry
Active
Our position
01
We prefer your company data stays on your own computer. On capable local hardware, Ollama, Obsidian, and OpenClaw run entirely free — your messages never leave your device.
02
If you want more powerful models, add your own API key at any time — OpenRouter, DeepInfra, Claude, GPT-4o, and others. The dashboard stays on your computer; only conversation content travels to your chosen provider.
03
The product works both ways. Start local, add cloud later — or the reverse. You can change your mind at any time.
Local-first
Structured comparison — what you get with local-first desktop software.
| Parameter | ClickSorcery | Typical cloud AI platform |
|---|---|---|
| Product type | Local desktop app (Windows & macOS) | Cloud SaaS (browser tab) |
| Architecture | Local-first — app runs on your machine | Hosted on vendor servers |
| Data location | Memory, vault, and settings on your device | Stored on third-party infrastructure |
| API keys | Encrypted in Vault on your device — never sent to ClickSorcery | Often held or proxied by the vendor |
| Model support | Ollama (local), OpenRouter, Groq, DeepInfra, Claude, GPT — your choice | Limited to vendor's bundled models |
| Memory layer | Markdown / Obsidian vault you own | Closed vendor database |
| Pricing | €79 one-time license + optional API usage you control | Monthly subscription + usage fees |
| Try before buy | 7-day full trial, no credit card | Varies — often card required or feature-limited free tier |
Jerry's brain
€79 is the app — not unlimited AI. Jerry runs on an LLM (language model) via your API key or local Ollama.
The AI that writes Jerry's replies — Llama, Claude, DeepSeek, …
Your account at Groq, OpenRouter, or DeepInfra — paste the key once into Vault.
~€5–15 light use · ~€10–40+ active daily (varies by model) · €0 with local Ollama (16 GB+ / GPU)
Choose your engine
Same Jerry. Different power source — and very different monthly bills.
Telegram and OpenClaw are free. You only pay when the AI brain runs in someone else's cloud.
| Component | Ollama (local home) | Claude / API (cloud brain) |
|---|---|---|
| Telegram bot | Free | Free |
| OpenClaw (software) | Free · open source | Free · open source |
| AI thinking (each message) | €0 API — your hardware + electricity | Pay per token (words sent + received) |
| Privacy | 100% on your machine | Prompts go to your API provider |
| Best for | 16 GB+ RAM or GPU — privacy, no per-message API bill | 8 GB laptops, smarter model, no hardware struggle |
Local: Telegram is free; thinking is free API-wise once Ollama actually runs. Cloud: every token counts — your question, Jerry's context, and the full reply. A short research chat can be ~€0.50; ~10 Claude questions with a small lookup has cost real users ~€5.
Add one API key. Jerry routes messages to the cloud — you pay per token, not per month. Memory and long replies count toward your bill too.
Monthly API cost — real-world, not wishful
| Groq / free tiers | Light chat, smart routing | ~€0–15 |
| Mixed cloud | Daily chat, some projects | €15–40 |
| Claude Sonnet daily | Research, long missions, premium models | €40–100+ |
Models Jerry can use
Why Cloud?
No flat fee — your API provider bills per token. Set a monthly cap so paid cloud never exceeds your limit.
Ollama on a machine that can actually run it (16 GB+ RAM or a GPU). OpenClaw talks to your local model — €0 API per message. Telegram stays free.
8 GB RAM without GPU: we don't recommend local Ollama — models often fail to respond and the PC fans spin with no answer.
If Jerry lives on this hardware
| 8 GB system RAM · no GPU | → Cloud Commander | Local Ollama not practical — machine struggles, Jerry may never reply |
| 16 GB system RAM · no GPU | qwen2.5:7b | Minimum local setup — slow CPU inference, but €0 API |
| 8 GB VRAM GPU | qwen2.5:7b | Jerry + up to 3 specialist agents |
| 16 GB VRAM | llama3.1:8b | 8 agents, daily workflows, routines |
| 24 GB+ VRAM | qwen2.5:32b+ | Full agent company — maximum local power |
Why Local?
GPU guide
| 4GB VRAM | Jerry mini |
| 8GB VRAM | 3 agents |
| 16GB VRAM | 8 agents |
| 24GB+ VRAM | Full company |
Est. cost: €0/month API + electricity
What you get on real hardware — we're honest about what 8 GB cannot do.
Typical 8 GB Windows laptop: once Ollama loads a model, the machine maxes out and Jerry often never answers. Use Cloud Commander instead — same Jerry, API key, no local inference.
You get
Limit: Heavy parallel missions can still slow an 8 GB PC. API costs depend on model — Groq/free tiers ~€0; Claude adds up. Use the in-app monthly cap so spend cannot exceed your limit.
Examples: ~€0 with light Groq chat · ~€5 for ~10 Claude questions + small research · €15–40+ if Sonnet is your daily driver
16 GB system RAM (CPU-only, slow) or a PC with a dedicated GPU. Ollama can finally load a small model and reply — qwen2.5:7b class.
You get
Limit: CPU-only is slow. Heavy automation and big agent teams need more hardware.
~€0 API · electricity only
Desktop with a real GPU. Jerry runs 7B–14B models comfortably — specialist agents can work in parallel.
You get
Limit: Not the full 40-agent company at max speed — but serious daily work.
~€0 API · electricity · optional cloud for peak tasks
Dedicated workstation or small server — hardware you buy for the company, not rent per message.
You get
Limit: Upfront hardware cost. You own the asset.
~€0 API · higher electricity
Honest default: 8 GB laptops → Cloud Commander. Local Ollama needs 16 GB+ RAM or a GPU — otherwise Jerry may never answer.
Disk space
Most laptops still ship with 256–512 GB — not multi-terabyte. Local AI models and automation stacks grow fast. Here is an honest breakdown before you install.
~1–3 GB total
Comfortable on a 256 GB ultrabook
~50–100+ GB possible
External SSD recommended for models, Docker/n8n, and vault growth
| Component | Size | Cost | Note |
|---|---|---|---|
| ClickSorcery app | ~127 MB download / ~740 MB installed | €79 one-time | Desktop installer |
| OpenClaw runtime | Varies | Free | Local shell — size depends on install |
| Obsidian vault | ~150–250 MB + growth | Free | Plain markdown memory — not a heavy local AI database |
| Ollama (engine only) | ~200–500 MB | Free | No models included |
| Local model (7–8B class) | 4–8 GB each | Free | Downloaded when you pull a model |
| Local model (30B+ class) | 20–40+ GB each | Free | Needs strong GPU or patience |
| Docker Desktop (n8n projects) | 6 GB base → 20–40 GB in use | Free | Only if you run automation projects |
| Cursor (optional editor) | 0.5–2 GB | Free / Pro | Not required to run ClickSorcery |
Local AI models are extra downloads on top of ClickSorcery itself — they are not bundled in the installer. One small model is enough to start; a full local company can fill a drive quickly.
If disk is tight (256 GB laptop), start with Cloud Commander — same product, minimal local storage, API costs instead of model files.
Pricing
€79 one-time Desktop Pro — early-bird / pre-sale for November–December 2026. After that we move toward Full Kwan (price not set yet — expect higher once the list below ships).
€79 = ClickSorcery desktop (one-time). Cloud AI is separate — ~€0–15/mo with Groq/free tiers, more with Claude Sonnet. Set a monthly cap in System & Budget; watch spend during your trial.
€79 one-time
one-time · Windows & macOS · pre-sale window
Desktop Pro is what you buy today. Full Kwan is the deeper human↔agent life layer we are building — nicknames, training diaries, calendars, hands-free, persona depth, and more. We do not publish the Full Kwan price yet; when these land, the product will cost more than the early-bird €79 window.
In progress / planned
No credit card · 7-day full access · €79 one-time early-bird license (Nov–Dec 2026)
FAQ
Yes. ClickSorcery is local desktop software for Windows and macOS. All API keys, Obsidian-based memory files, agent settings, and workflows are stored 100% on your device. Keys are encrypted in the Electron main process (Vault) — not in the browser bundle and not on ClickSorcery servers. We have no technical access to your keys, conversations, or vault contents. In Cloud mode, message content goes directly from your device to the API provider you choose (Groq, OpenRouter, Anthropic, etc.) — not to ClickSorcery. Details: https://clicksorcery.com/privacy
ClickSorcery is a trademarked desktop app by Wicmath OÜ (Estonia). Start with Jerry — one AI for work, home, learning, fitness, and everyday tasks on your computer. When you need more, the same app includes 40 specialist agents and verified workflows (local-first on Windows and macOS). Not a generic cloud chatbot subscription — your keys, your vault, your machine. €79 one-time; optional cloud AI via your own API keys or fully local Ollama mode.
56 guided tasks and 86 project templates in Desktop Pro are two different things — and you do not need to finish all of them. Guided tasks = a 14-level interactive onboarding railway: you configure Jerry's brain, voice, Vault, Approvals, Telegram, and your agent team on your real machine (not theory in a PDF). Finishing every task makes you a power user; it does not unlock a hidden tier or 86 autonomous services. Project templates = an executive menu (buffet), not 86 mandatory quests. Need a launch plan, content calendar, ebook outline, or contract review today? Pick one template → Start with Jerry → structured plan, drafts, and next steps (often via Approvals). Many templates are LLM-only; others need your keys, Telegram, Google, n8n, or DIY hardware — the app says so upfront. You stay Chairman: you open the door (install, key, account); agents prepare analysis and drafts; you review and Approve before anything live. Full project list: https://clicksorcery.com/projects — agents: https://clicksorcery.com/agents
No. Setup wizard + Jerry guides everything.
€79 one-time for ClickSorcery Desktop Pro (Windows & macOS). Cloud API costs are separate — light use ~€5–15/month; active daily often ~€10–40+ depending on model and chat length. Local Ollama: €0 API if your hardware qualifies (16 GB+ RAM or GPU). Budget monitor shows your real spend during the trial.
Cloud Commander needs ~1–3 GB total disk (app + vault, no large model downloads). Local Supercomputer plus automation projects can reach ~50–100+ GB — plan an external SSD for Ollama models and Docker/n8n. Local AI models are extra downloads, not bundled in the installer. Details: https://clicksorcery.com/#disk-space
No! Start on your regular 8 GB laptop with Cloud Commander (~€0–15/mo with Groq, more with premium models). Set the in-app monthly cap so costs stay predictable. When ClickSorcery pays off for you, invest in local hardware later — same product, different energy source.
No — ClickSorcery uses smart model routing. When Brain is Auto, agents try Groq, then DeepInfra, then OpenRouter by role — cheap tiers first, premium models on failover. You can also pin a model per agent in Settings.
Desktop Pro includes Mobile Jerry — a phone PWA (QR pairing, optional Tailscale for away-from-home). Push-to-talk with Jerry or any agent; optional phone or home LAN camera vision. Desktop must stay running at home. Guided onboarding: Task #2 voice → Task #3 phone. Telegram is an alternative for quick messages. Full Kwan native app store listing is not sold today.
For a hands-free workout or run: use Mobile Jerry with Bluetooth earbuds, select Fitness Coach (or another Personal Branch coach), and use push-to-talk. Your home desktop runs the hub — this is not a phone-only cloud coach. Each reply is one turn (speak, release, listen); we do not ship sub-second full-duplex streaming (Deepgram live / OpenAI Realtime) in Desktop Pro today. Live coach check-ins use the fast Groq/voice stack when a Vault key is present so a pace cue is not stuck behind a two-minute Sonnet wait. STT and TTS use your Vault API keys (OpenAI/Groq + optional Cartesia/ElevenLabs). Per-coach voice presets: Settings → Personal coaches. Not medical advice — stop if you feel unwell.
Jerry voice is three steps: listen (STT) → think (LLM) → speak (TTS). You can keep thinking at €0 on local Ollama. Free speech: Edge-TTS (Microsoft neural voices, no API key, Estonian Anu/Kert, needs internet) or Piper (local CPU, offline after an ~80 MB install; official voices are English, not Estonian). The old OS system voice is still there for a robotic trial. Paid cloud TTS (OpenAI, ElevenLabs, Cartesia) stays optional in Vault. F5/XTTS-style GPU cloners are not wired — they would fight Ollama for VRAM. Listening: Groq Whisper (free tier with a Vault key) is still the best STT. Full tables: Settings → Voice → (i) in the desktop app.
Jerry reply speed is not a single number. It depends on (1) the model — Groq/Haiku-class is seconds, Sonnet-class or a 70B local model is tens of seconds; (2) streaming — words should appear as they are generated, not only when the full answer is done; (3) extras — vault search (RAG), web tools, and a long chat history add time; (4) OpenClaw warm-up on first launch, especially on an older PC. What you can do: add a Groq key in Vault, leave Brain on Auto, keep everyday questions short, start a new chat when the thread is huge, turn RAG off when you do not need vault search, and use the chat readiness bar to start/restart OpenClaw. Live coaches (fitness, nutrition, time, language, speak-up) stay on the fast stack for check-ins. What you cannot skip: a real plan, strategy, or tool call cannot match “hi” in ChatGPT. We will not fake instant replies on a full mission plan or a large local model. Details: https://clicksorcery.com/#reply-speed
No — and that's a feature. Jerry routes each agent to the best model for their job: the CFO reasons with DeepSeek-R1, the CTO codes with Qwen-2.5-Coder, the researcher searches with Perplexity Sonar, and Jerry himself runs on Claude. One API key, 400+ models, automatically assigned.
Yes with Cloud Commander (API key). Local Ollama on 8 GB usually fails — PC maxes out and Jerry often never answers. Use cloud on 8 GB; use local on 16 GB+ or a GPU.
Yes — Jerry + 3 agents. 16 GB VRAM → 8 agents. 24 GB+ for full local company.
Local Ollama mode: 100% on your machine — nothing sent to any cloud. Cloud mode: your prompts go only to the API provider you connect (not ClickSorcery). Vault, memory files, and keys never leave your device. See /privacy.
Yes. Export your AI Company (.zip) from Settings, install ClickSorcery on the new machine, unlock with your license, and restore. One active machine at a time.
ClickSorcery offers a 7-day free trial with full access — no credit card required. Use the trial to evaluate everything before you buy. All license purchases after checkout are final; we do not offer refunds or money-back guarantees. Questions before buying: help@clicksorcery.com
40 profiles in the app. Jerry (CEO) + 5 C-suite + Layer 0 agents + department specialists + personal coaches (e.g. language, fitness, nutrition) — each with its own role and brain routing. You enable agents as you need them; Jerry coordinates the team. Named list: https://clicksorcery.com/agents
When you approve a routine, ClickSorcery runs it on a schedule while the desktop app is open on your PC — there is no cloud cron that works while your computer is off. Heartbeat checks run on a similar timer. Full Kwan may add an always-on home hub later; Desktop Pro today is local-first: PC on, app running.
No automatic 2 a.m. job. Vault Steward uses a local Ollama model to propose note structure; you approve before anything is written. Use the home Vault Steward panel or the chat prefix “structure this for Obsidian vault: …”. The ingestion queue is processed when you choose — not silently overnight.
Jerry uses Obsidian-based memory — your projects, preferences, and past conversations persist across sessions. He doesn't start fresh every time.
Yes — that's ClickSorcery's core promise. Jerry guides setup via the Magic Installer. You talk in plain language; no prompt engineering or developer skills required. The product ships with 40 pre-configured agents ready from day one.
Yes. Telegram is included in Desktop Pro: connect your bot in Settings, chat with Jerry from your phone, send voice notes and get audio replies. Your PC must be running for inbound messages; outbound alerts work when configured.
Yes — Local Supercomputer mode uses Ollama on your Windows or macOS desktop. 100% private, no per-message API cost. Requires 16 GB+ RAM or a dedicated GPU. On 8 GB laptops, use Cloud Commander mode with your own API key instead.
ClickSorcery is an AI Company OS: 40 agents led by Jerry (AI CEO) for finance, marketing, ops, research, and support — 20 core verified workflows + 66 templates, 56 guided tasks, 65 tools. One desktop app, one conversation with Jerry, work delegated to specialists automatically.
After purchase, you can ask Jerry directly — he knows everything about ClickSorcery and answers instantly.
We reply from help@clicksorcery.com.