01
Web research
Reads the open web. No third-party tracking.
A working operating system
for the principal who runs the work.
It runs on a server in your name. It remembers everything. It does not require management.
01 — The condition
The operator becomes the bottleneck.
The operation becomes fragile.
Operational knowledge collects in the principal. Decisions wait. The business grows. The principal does not scale.
02 — The premise
Han AI is not a tool.
It is the layer beneath the tools.
A memory-bearing agent on the principal's own server. It uses real tools — web, email, documents, browser, code, schedule — to do real work.
03 — Distinction
Software grants access.
Han AI installs an operating layer.
The distinction is whether the system can remember, execute, govern, and improve real work.
Layer
Common AI tools
Han AI
Memory
Context scattered across chats, tools, and people.
Operational knowledge becomes durable memory.
Execution
The system suggests, summarizes, or waits.
Recurring workflows run on schedules, queues, and approvals.
Governance
Authority is unclear once work moves between tools.
Approvals, escalation, and audit hold authority with the principal.
Learning
The same issues repeat because improvements stay informal.
Improvements are proposed from evidence and approved before behavior changes.
04 — Method
The work moves.
The tools are how.
A question in Telegram becomes a published post, a sent letter, a paid invoice, a delivered brief. Nine tools sit underneath. All run on your server. No third-party API in the critical path.
02
Documents
Opens PDFs, Word, ODT. Pulls the substance.
03
OCR
Reads scanned images and screenshots, including Khmer. Local.
04
Sends from your own SMTP. For high-trust correspondence.
05
Scheduled execution
Weekly, monthly, quarterly jobs. The schedule persists.
06
Sandboxed computation
For when the number must be exact.
07
Browser automation
Drives Chromium. Reads JavaScript-rendered pages.
08
Persistent memory
Recalls documents by intent, not by filename.
09
Premium documents
Publication-grade PDFs. Learns the styles you prefer.
05 — Terms
Three tiers.
The price is the price.
Sold as an operational deployment, not as software access. Onboarding is paid at signing. The fee is billed monthly.
Operator
$1,500
per month
$2,500 onboarding
20M tokens / month
A solo founder or a small team. One business. Real operational complexity.
Founding tenant
Business
$2,900
per month
$2,500 onboarding
40M tokens / month
Five to fifteen staff. Multiple revenue streams. Agency or operator-led.
Multi-brand
$5,900
per month
$5,000 onboarding
Unlimited tokens
Fifteen-plus staff. Multi-brand or multi-business operator.
Founding tenants.
The first three Business-tier clients receive a twelve-month price lock and a direct line to Han for the first ninety days. The engagement is used as a case study — named or anonymous at the client's choice. After three slots, full pricing.
A quarterly cap on new clients holds operational maturity above growth. When full, prospects join the next quarter's waitlist.
06 — For principals
A separate line,
by invitation.
An invitation-only line for principals, family offices, and multi-business operators whose privacy and discretion requirements exceed the Business line.
Configurations from $5,000 per month.
Write to private@hanai.systems.
07 — What evaluators ask
Where does the data live?
On your own Vultr VPS, in the region chosen at provisioning. Plus your own Airtable base. No central platform server.
Which model does Han AI run?
Foundation models from Anthropic and OpenAI, selected per task by the runtime. The AI Use Disclosure is part of the contract pack at signing.
What happens at exit?
The VPS, the Telegram bot, the Airtable base, the captured profile, the conversation history — all belong to you. Handed over. No lock-in.
How are approvals held?
Han AI never sends an email, makes a payment, or contacts a third party without your sign-off. The boundary widens or narrows on request.
PCI, GDPR, HIPAA — alignment?
Each engagement is contracted under MSA and DPA. For regulated workloads, the specific framework is raised at discovery and quoted accordingly.
Why a VPS instead of the cloud?
A real server with shell access is what permits the system to act — open documents, run code, drive a browser, schedule jobs. Without it, this would be a chat product.
08 — Request a deployment
Request a deployment.
We will respond within two business days. Or write directly to hello@hanai.systems.
Received.
Thank you. We will respond within two business days.