What Han AI does.

An agent fleet that runs the recurring operational load of a service business — on the principal's own server, under their authority.

Han AI is a memory-bearing operating system for service businesses. It holds the operational knowledge of the business and executes the recurring load — scheduling, delegation, reporting, publishing, financial tracking, and governance — without requiring the principal to initiate each task. Authority stays with the principal at every step.

The work Han AI coordinates

The system runs eight categories of operational work:

• Production scheduling. Client cycles, shoot calendars, delivery timelines — planned and tracked without manual setup each cycle.
• Task delegation. Assignments surface at the right moment and are tracked to completion.
• Client reporting. Status summaries, delivery confirmations, and status updates prepared and queued for principal review.
• Content publishing. Scheduled publishing across platforms, on the cadence set by the principal.
• Internal briefings. Morning and evening operational briefings delivered automatically — what is due, what is at risk, what requires a decision.
• Financial tracking. Payment follow-up, revenue tracking, and recurring cost monitoring surfaced to the principal.
• Memory accumulation. Everything the system learns about how the business operates is written to durable memory and recalled in context.
• Governance. Approvals, escalations, and the audit boundary — nothing crosses to a client, a third party, or an external service without sign-off.

How the coordination works — OpenClaw

Han AI's agent fleet runs on OpenClaw, an AI agent framework that manages the task queue, execution routing, and governance boundary beneath Han AI's operational surface.

The agent fleet covers the operational domains above. Each agent is scoped to its function, reads only its own history, and writes findings to a shared intelligence layer that other agents can draw on in context. The coordination layer — not any single agent — is what makes the system an operating system rather than a collection of tools.

The nine tool layers underneath

Underneath the agents sit nine execution tools that do the actual work:

• 01Web research — reads the open web without third-party tracking
• 02Document processing — opens PDF, Word, and ODT files
• 03OCR — reads scanned images and screenshots including Khmer script, processed locally
• 04Email — sends from the principal's own SMTP for high-trust correspondence
• 05Scheduled execution — weekly, monthly, and quarterly jobs that persist across restarts
• 06Sandboxed computation — exact numeric work in an isolated environment
• 07Browser automation — drives Chromium and reads JavaScript-rendered pages
• 08Persistent semantic memory — recalls documents by intent, not by filename
• 09Premium document generation — publication-grade PDFs that learn the principal's preferred styles

The governance boundary

The system proposes; the principal approves. Han AI never sends an email, makes a payment, or contacts a third party without the principal's explicit sign-off. The boundary of autonomous action can be widened or narrowed on the principal's instruction — but it is always present.

Internal state — briefings, memory, agent outputs — is autonomous. Outward action requires a gate.

Where it runs

Each deployment runs on the principal's own VPS, not on a shared platform. The Telegram interface, the Airtable base, and the operational profile all belong to the client. At exit, everything is handed over — server, bot, base, memory, conversation history. No lock-in.

The live proof environment

Han AI was built and is operated by Han Khim, Founder of Han Studios, a creative production and marketing company in Phnom Penh, Cambodia. Han Studios runs all of its studio operations on the same system offered to clients — production scheduling, client reporting, content publishing, financial tracking, and governance. The system is not theoretical.

Frequently asked

Request a deployment →