Setup
What the platform runs on, and what it costs
The current stack plus what each piece does, then the monthly cost today and the projected cost with every automation in the Future Solutions catalogue live. The server cost is fixed. The only thing that scales is the AI bill, and even that stays small.
01Tech stack
Eight pieces, all open source or managed services with generous free tiers. The whole stack lives on one Hetzner server.
Hosting
Hetzner CPX31
The server. Four CPU cores, eight gigabytes of RAM, 160 GB of disk. Hosts everything: the API, the database, the frontend, n8n. Fixed cost, does not change as more automations are added.
Docker Compose
Each piece of the stack runs in its own container. To deploy a change, the server pulls the latest code from GitHub and rebuilds. Done in under two minutes.
Backend
FastAPI (Python)
The API. Validates form input, talks to Gemini, sends emails, posts to Telegram and Slack, writes to the database, runs background polling on the Google Sheet.
PostgreSQL
Stores every placement, every activity log, every runtime setting. When you ask the dashboard a question, this is what answers.
Channels and integrations
Google Gemini
The model that writes the client emails. Cheap, fast, good at short professional copy. Currently using Gemini 2.0 Flash.
Gmail SMTP
Sends the client emails. Uses an app password on a normal Gmail account, no special API setup.
Telegram Bot
Pings the worker when they are assigned. Free, fast, easy to demo. The bot can also notify the team or post to a channel.
Slack Incoming Webhook
Posts a one-line note to the team channel for every placement, and an alert when a submission is rejected.
Google Sheets API
Reads the intake sheet every 30 seconds for new rows and writes placements back to it so the sheet stays in sync.
n8n
An optional second engine that can run the same workflow visually. Self-hosted alongside the backend.
Frontend
Next.js 15 + shadcn + Tailwind
The dashboard you are looking at right now. Server-rendered for speed, client-side for interactivity. The same codebase serves all eight pages plus the walkthrough.
02Monthly costing
One number to remember
The server is fixed at twenty US dollars per month. Everything else scales with how much the system is used, and even at high volume the API bill stays under five dollars.
Current setup — what is live today
| Item | What it is | Monthly |
|---|---|---|
| Hetzner CPX31 | 4 vCPU · 8 GB RAM · 160 GB | $20.00 |
| Google Gemini 2.0 Flash | ~3,000 emails/mo @ $0.0002 | $0.60 |
| Gmail SMTP | Free tier (500 emails/day limit) | $0.00 |
| Telegram Bot API | Always free | $0.00 |
| Slack Incoming Webhook | Always free | $0.00 |
| Google Sheets API | Free under quotas | $0.00 |
| GitHub (private repo) | Free | $0.00 |
| Total | Per month, current usage | $20.60 |
With all 15 automations live
| Item | What it is | Monthly |
|---|---|---|
| Hetzner CPX31 | Same server, all automations on it | $20.00 |
| Gemini — placement emails | ~3,000 calls | $0.60 |
| Gemini — client lead replies | ~500 leads/mo | $0.10 |
| Gemini — pre-screen interviews | ~200 worker chats | $0.20 |
| Gemini — smart matching | ~1,500 matches | $0.30 |
| Gemini — followups & retention | ~3,000 followup messages | $0.60 |
| Gemini — invoicing notes | ~3,000 invoice cover lines | $0.60 |
| Gemini — weekly digests | ~4 long-form summaries | $0.10 |
| Gemini — other calls | Miscellaneous (validation, KYC reads, embeddings) | $1.50 |
| Gmail, Telegram, Slack, Sheets | Same as today, still free | $0.00 |
| Optional Twilio (WhatsApp/SMS) | Pay per message, only if enabled | varies |
| Optional Stripe (invoicing) | 2.9% + 30¢ on actual payments only | n/a |
| Total | Per month, all automations active | $24.50 all 15 live |
That is roughly the price of two cups of coffee. The reason it stays so cheap is that every automation uses the same server and most of them use the same Gemini model. Adding a new automation does not add a new subscription, it adds a few hundred API calls to the existing one.
What changes if usage spikes
More AI calls
Gemini is billed per call at roughly $0.0002 per email-sized response. Going from 3,000 to 30,000 placements a month adds about $5.40 to the bill.
More server load
The CPX31 handles many thousands of placements per month before it needs upgrading. The next tier up costs about $35 per month.
What this means for adding new automations
Each new automation in the Future Solutions catalogue is a one to four week engineering build, but the monthly cost only goes up by cents. The server does not change. Nothing new needs to be subscribed to. Everything stays on the same dashboard you are using today.