AI that works for your business — on your own hardware
Not a chatbot bolted onto your website. An agent that lives on a machine you control, remembers your business, reaches you on WhatsApp or Telegram, and quietly gets work done overnight. Your data never leaves the building.
Yes. Fentech builds AI agents that run on your own machine or a VPS it manages, so the agent, its memory and its history never leave your control — no per-seat subscription and no telemetry. Typical jobs are inbox triage, draft quotes, monitoring and recurring reports. Scoping conversations are free.
The jobs that eat your evenings, done before you're up
These aren't hypotheticals. They're the kinds of automation Fentech runs on its own business every day — the weekly blog drafts on this site come out of exactly this setup.
Inbox triage
Reads what arrived overnight, sorts the noise from the money, drafts replies to the routine ones, and tells you which four actually need a human.
Quotes & follow-ups
Turns an enquiry into a draft quote in your format, and chases the ones that went quiet — politely, on a schedule, without you remembering.
Morning briefings
What sold, what's due, what broke, what needs signing — pulled from your systems and delivered to WhatsApp before your first coffee.
Monitoring that tells you first
Watches your servers, sites, certificates and backups. When something's wrong you hear it from the agent, not from a customer.
Research & watching
Competitor prices, planning applications, tenders, industry news — collected, summarised and filtered to what matters to you.
The boring recurring jobs
Reports compiled, files renamed and filed, data moved between systems that don't talk to each other. Every day, correctly, without complaint.
Count the hours before you spend the money
The honest test for any automation is simple: how long does the job take, how often does it happen, and what is an hour of your time worth? Drag the sliders.
Count the interruption too — a five-minute job that breaks your concentration ten times a day costs more than fifty minutes.
Worth a conversation.
Talk about this jobYour data, your hardware, your rules
Cloud AI subscription
- Your emails, files and customer data sent to a third party's servers
- Per user, per month, forever — rising as you grow
- Terms of service you didn't read decide what happens to your data
- Features removed or prices raised on their schedule
- Awkward answers when a client asks where their data went
Self-hosted agent
- Runs on a machine you own — in your office or on a server you control
- Memory, skills and history stored locally, not in a vendor cloud
- Open-source foundations you can audit line by line
- No telemetry, no tracking, no data collection
- A straight answer to the GDPR question: it never left
The models you choose can run fully on-premise too. For some businesses the right answer is a capable hosted model behind a locked-down gateway; for others nothing should ever leave the room. Both are built here.
The stack, in plain English
Open-source agent platforms chosen because they run on your hardware, keep their memory locally, and don't phone home. Fentech runs both on its own infrastructure.
Hermes Agent
Open-source, MIT-licensed agent from Nous Research. Persistent memory that improves the longer it runs, 40+ built-in skills that it extends by writing its own, a gateway into Telegram, Discord, Slack, WhatsApp and Signal, a built-in scheduler for unattended nightly jobs, parallel sub-agents, and full browser control. Runs local models on-premise or any compatible API.
OpenClaw
Open-source assistant that runs on your own machine and works from the chat apps you already use — WhatsApp, Telegram, Slack, Discord, iMessage, Signal. Organises inboxes, sends email, manages calendars, controls a browser, works with Gmail and GitHub, and connects to Claude, GPT or local models. Desktop apps for Windows, macOS and Linux; a community running fleets of agents across multiple machines.
Models
- Claude — reasoning, writing, review
- GPT — general tasks, tool use
- Local open models via vLLM — fully on-premise
- OpenRouter — 200+ models behind one key
Reach
- WhatsApp · Telegram
- Slack · Discord · Signal
- Email and terminal
- Start on one, continue on another
Runs on
- A box in your office
- Your existing server
- A VPS Fentech manages for you
- Docker, hardened and isolated
Resilience
- Multi-provider failover — one API having a bad day doesn't stop the work
- Encrypted off-site backup of the agent's memory
- Monitoring on the monitor
Two ways to work together
- Discovery: what eats your time, what's safe to hand over
- Agent installed on your hardware or a managed VPS
- Connected to your email, calendar, chat and business systems
- First automations built, tested and running
- Guardrails: what it may do alone, what needs your say-so
- Written handover — you own it outright
- Everything in the setup project
- New automations added as your business changes
- Models, updates and security patches kept current
- Monitoring, backups and failover looked after
- A person to ring when you want it to do something new
- Monthly summary of what it did and what it saved you
Agentic AI, answered plainly
What would you hand over first?
Describe the job that eats your time. You'll get an honest view of whether an agent should be doing it, what it would take, and a fixed price — no obligation.