ClickSorcery
ClickSorcery · Journey

Every company
starts somewhere

From an 8 GB laptop with a €50 API budget to your own full AI company — running locally, 24/7, without cloud bills.

5 stages
on the path
€0
API bill at the end
70B+
models locally
€0.10–€0.35
kWh · electricity by region
💻
Stage 01 · Start
Curious
"Not sure yet — but I want to try"
€50+
month · your API budget
€0 hardware

ClickSorcery works great within a normal API budget. Set your cap — €50, €100, €500 or more — and the system optimizes tasks inside it. Your budget, your pace.

Your monthly budget: €100

5–8 agents · meetings + email · several projects

💻
Your existing PC / laptop
8–16 GB RAM · whatever you already own
Zero hardware spend
Claude SonnetGPT-4oDeepSeek V3 APIClickSorcery €79 one-time
First automation workflows
Meeting notes with AI
Draft email replies
Project planning with Jerry
Marketing content drafts
First "wow" moment — hours back
Show PC variant
Same PC / laptop (same stage)
8–16 GB RAM · cloud API
€0 hardware · API per your budget

No PC swap needed — stay in cloud mode.

When to move on? If your monthly API bill keeps crossing ~€80, local hardware pays for itself. ClickSorcery shows you that moment — automatically.

💡
Upgrade when: You want privacy, offline work, or your API bill keeps climbing past ~€80/month.
Investment: ~€800–1,200Mac Mini M4 · 16/24 GB
🖥️
Stage 02 · First local
Believer
"It works. And it's private."
€13–26
month · electricity (~0.09 kW)
+ ~€800–1,200 one-time
🖥️
Mac Mini M4 · 16 GB or 24 GB
10-core CPU · 10-core GPU · Ollama locally
First step
Llama 3.2 8BMistral 7BGemma 3 12BPhi-4 14B (slow)70B ✗ not yet
Local Jerry 24/7 (slower but private)
Data stays at home
Works offline without internet
8–14B models · fine for lighter tasks
API bill nearly disappears
Learn Ollama and prompting

32B models and a full department → see Stage 3 (Mac Mini M4 Pro 64 GB).

Show PC variant
1× RTX 4090 workstation
24 GB VRAM · models up to ~13B
~€4,000–5,000

On PC you run smaller models for similar money — no unified memory.

💡
Upgrade when: 8B models feel too weak for your work. You want 30B+ intelligence locally.
Investment: ~€2,400Mac Mini M4 Pro · 64 GB
⚡
Stage 03 · Real power
Operator
"One department. Fully automatic. 24/7."
€19–40
month · electricity (0.15 kW)
+ €2,400 one-time
⚡
Mac Mini M4 Pro · 64 GB Unified Memory
12-core CPU · 20-core GPU · OLLAMA_NUM_PARALLEL=4
Real 30B
Qwen 2.5 32BDeepSeek Coder 32BMistral Small 24BGemma 3 27BPhi-4 14B · fast
30B model ≈ GPT-3.5 class locally
One fully automatic department (Sales / Dev / Marketing)
Jerry runs 24/7 locally
4 parallel agent calls
Code from 32B coder model
API bill: €0
All data at home
Break-even vs API: <1 month
Show PC variant
Dual RTX 4090 tower
48 GB VRAM total · 32B split across GPUs
~€8,000–10,000

32B is possible on PC but slower and fussier than Mac Mini 64 GB.

💡
Upgrade when: One department is automated and you want another — or you need 70B capability, not just an assistant.
Investment: €2,400 × nAdd Mac Minis department by department
🏗️
Stage 04 · Multi-department
Builder
"Three departments running at once"
€38–80
month · electricity (0.30–0.45 kW)
+ €4,800–7,200 investment
🏗️
2–3× Mac Mini M4 Pro 64 GB
Network: Wi-Fi 6E / Ethernet · Ollama OLLAMA_HOST=0.0.0.0
Mini cluster
Qwen 2.5 32B × 3DeepSeek Coder 32BEach machine its own model~8–12 parallel calls
Marketing Mini: content, social, ads
Sales Mini: leads, CRM, proposals
Dev Mini: code, fixes, PR review
All run in parallel
Jerry coordinates every Mac
Each Mini = its own department
Scale as you need
API: €0/month forever
Show PC variant
Dept PCs · RTX 5090 32 GB
Independent machines · 32B per node
€3,500–3,700 each

Mac still wins quiet ops; PC scales with standard workstations — no NVLink farm.

💡
Upgrade when: Ready for the full cluster — a Boss machine running 70B models and coordinating every Mini.
Investment: M3 Ultra BossMac Studio M3 Ultra Boss (~192 GB tier)
🏆
Stage 05 · Full stack
Dream Stack
"Full company AI OS. Private. Powerful. Yours."
€71–238
month · electricity (0.66 kW)
~€20,000 total · vs API: pays back in <30 days
⚡
OPENCLAW BOSS — Mac Studio M3 Ultra (~192 GB)
24-core CPU · 76-core GPU · OLLAMA_NUM_PARALLEL=8
apple.com/ee
🖥️
5× Mac Mini M4 Pro 64 GB — Department machines
Marketing · Sales · Dev · Finance · Ops · 10GbE network
5 × €2,400 = €12,000
Qwen 2.5 72B (Boss)DeepSeek-R1 70B (Boss)Qwen 2.5 32B × 5DeepSeek Coder 32B~30 parallel callsAll local
Jerry orchestrates 33 agents live
Boss decides, Minis execute
70B model ≈ GPT-4 class locally
5 departments in parallel 24/7
Vault = Obsidian memory at home
€0 API bill forever
Full privacy — zero cloud
Save vs API: €676k+ per year (at scale)
Show PC variant
RTX PRO 6000 Boss + 5× RTX 5090
96 GB Boss · 32 GB/dept · ~4.6 kW · ~50–60 dB
~€30,000

Honest PC path: one 96 GB Boss GPU, five department PCs — faster Boss bandwidth, higher power bill.

🏆
You reached the end
ClickSorcery at full power. 33 agents. 5 departments. Fully automatic. Private. Up to 50× cheaper than cloud API — break-even within the first month.
The whole journey at a glance
StageHardwareModelsCost / monthInvestmentAPI bill
Curious
Stage 01
Existing laptopClaude / GPT-4o / Sonnet€50+€0Running
Believer
Stage 02
Mac Mini M4 16/24 GB8B–14B locally€13–26~€800–1,200Nearly €0
Operator
Stage 03
Mac Mini M4 Pro 64 GBQwen 32B · Coder 32B€19–40€2,400€0
Builder
Stage 04
2–3× Mac Mini M4 Pro 64 GBQwen 32B × 3 · parallel€38–80€4,800–7,200€0
Dream Stack
Stage 05
M3 Ultra Boss + 5× Mini 64GBQwen 72B + 32B × 5€71–238~€20,000€0

Where are you today?

Start at stage one. Each step simplifies your life and moves euros from API bills back into your pocket.

Start with ClickSorcery →