What is UpServe?

UpServe is a platform for hiring an AI employee and handing work to it. Unlike a chatbot, an agent picks the right tools and runs the task to completion after a single instruction — and the next day it clocks in on its own, still remembering the job.

You can build and grow one yourself, or have our team build it for you and keep looking after it (see Build it yourself, or hand it off? below).

What can you do with it?

Repetitive work with a defined standard fits best. Here is what people actually hand off:

• 🧾 Partner order intake → ERP — Partners place orders like a chat through a single link; the agent asks back for anything missing and logs the order in eCount ERP
• 📦 Inventory checks & reorder lists — Checks ERP inventory at the same time, by the same standard, every day, and compiles the shortage reorder list
• 🗓️ Booking intake → calendar & ad management — Logs Naver Smart Place bookings into Google Calendar, reviews the ad report, and adjusts underperforming keywords
• 📄 Repetitive document review & reporting — Catches missing items and inconsistencies by the same standard and writes the review report. You only do the final check
• 🛒 Product DB management — Daily data entry like new product listings and stock updates on your storefront
• 📰 Morning briefings — Checks markets, news, or competitor pricing on its own and flags only the changes that matter
• 🎬 Content production — Plans and produces one YouTube Short a day in your channel’s tone, ready to upload

Services without a dedicated integration (Smart Store, Naver Reservations, and the like) are handled by the agent driving a real browser itself.

How is it different from other AI tools?

Let’s be honest: “the AI searches the web, schedules tasks, and remembers your conversations” isn’t special anymore. ChatGPT and Gemini now use tools, run scheduled tasks, and learn from memory too. So lining them up on those features misses the point.

The real difference isn’t a feature checklist — it’s what actually does the work.

• A chatbot is a single conversation that moves only when you speak to it. Close the window and it stops; reopen it and you’re just continuing a chat.
• A UpServe agent is a worker you own — one that exists on its own and runs on its own. Close your laptop, or fall asleep, and it still wakes up in the cloud to do its job, triggered by a schedule, a webhook, an external event, or a call from another agent.

In one line: a chatbot is a tool you use; a UpServe agent is an AI employee that works for you.

One picture

[Chatbot]
   You ──message──▶ Chat window ──answer──▶ You
        (exists only while you keep the window open)


[UpServe agent]  —  a worker that runs on its own, even when you're away

   What wakes it          What it does
   ─────────────          ────────────
   ⏰ a schedule           searches the web, drives a browser,
   🔔 webhook / event ──▶  writes files, runs code,
   💬 another agent        runs risky actions only after approval,
   🙋 your instruction     delivers the result ──▶ 📱 You

   (keeps its own workspace · accumulated memory · connected accounts)
   (keeps running in the cloud even while you sleep)

And yes, it works deeply from a single instruction too. Given one line like “Show me today’s AI/semiconductor briefing”, the assistant runs web search several times (see web_search (13×) below), manages the work steps as a checklist, and delivers an organized result — sources attached.

Build it yourself, or hand it off?

There are two ways to use UpServe.

The managed track is aimed at businesses of 1–5 people and supports tax invoices and bank transfer. These docs are written for self-serve — for what changes on a managed contract (an “Organization usage” screen in place of a SU balance, the contracted duties list, and so on), see the managed section of SU & Billing.

What about the server?

You don’t need one. No Mac mini running 24/7, no separate cloud VM to rent. UpServe handles hosting, execution, and scheduling for you.

• 🖥️ No local machine required — Close your laptop, lose Wi-Fi, the agent keeps running in the cloud
• ⏰ Always-on — A 3 AM cron schedule fires on time, regardless of your device state (on the Free plan, going 7+ days without using your account at all auto-pauses your agents, and scheduled runs stop with them; paid plans and agents in a managed organization are exempt from this auto-pause)
• 🔧 Zero infra ops — No Docker, Kubernetes, or server patching to worry about

A dedicated virtual workspace per agent

Every agent gets an isolated sandbox as its private workspace. Code execution, file operations, and browser automation all happen inside it — agents never interfere with each other.

Inside that environment, the agent can run Python scripts through a shell (and install other runtimes on demand if needed), drive a real browser, and persist working files between turns.

Which models can it use?

UpServe brings together models from Anthropic, OpenAI, Google, MiniMax, and more — and you can also register your own self-hosted endpoints. What you choose is not a model but one of two tiers.

If you do not pick a tier, Standard is applied. After creation, you can change the tier anytime from each agent’s Settings tab. Both tiers are available on every plan, including Free. Note that Ultra spends far more SU (roughly 10×), so move only the agents that need it.

UpServe decides which AI model backs each tier. You pick the tier; what runs behind it is our job. When a better model ships we swap it in tier by tier, so your agents improve without you changing a setting.

What can you explore next?

Here is a quick look at the major features waiting for you.

Next steps

• Core Concepts — Get clear on what agents, tools, skills, and teams are
• Build Your First Agent — Build something in 5 minutes
• Mobile App — Manage agents from your iPhone or Android phone
• Approvals — How Human-in-the-Loop works
• Managing Agents — Memory cleanup, pausing, reviewing run costs, and more