Skip to Content
Get StartedHire Your First Employee

Hire Your First Employee

Follow this guide to end up with a “news assistant” that organizes the market trends you care about every morning. Takes about 5 minutes.

In UpServe you don’t write behavior instructions by hand. Press hire and someone new starts right away — and they ask you what the job is themselves. You just answer like you would a new colleague, and they write their own name, mission, and personality from that conversation.

Prerequisites

  • An UpServe account (sign up at upserve.app )
  • 1–3 keywords you care about (e.g. “AI startups”, “semiconductor market”)

When you first sign up, the manager panel opens once on its own and walks you through the same steps as this guide, as a checklist. Feel free to close it — reopen the manager later and a “Back to onboarding” button appears at the top of the panel to pick up where you left off.

Step 1: Press hire

  1. Open the Office page 
  2. Click the + Hire button at the top of the screen (the sidebar’s +Hire goes to the same place. On mobile, tap + at the top right → Hire)
  3. Confirm on the dialog — on web it reads “Hire a new employee?” with a Confirm button; in the mobile apps the button says Hire

That card asks one thing and one thing only. No name, no job, no specs — the new employee asks about all of that themselves.

Confirming creates the employee immediately and opens their chat. There is no waiting screen.

Step 2: The new hire asks first

Your new employee opens by asking what job they’re taking on. That conversation is their onboarding.

New hire — Hi, I started today. Could you tell me what I’ll be working on?

You — Keep an eye on the markets I care about and brief me every morning. Mostly AI startups and semiconductors.

New hire — What time works best for the morning briefing?

You — Before I get in, around 8.

New hire — What should I go by?

You — Morning Brief works.

You don’t have to answer every question. Say “you decide” and they’ll fill it in from the conversation.

What the new hire decides for themselves

ItemWhere it comes from
NameDerived from the conversation (e.g. Morning Brief) — or whatever you tell them
Mission & descriptionWritten by them during onboarding
Personality & toneWritten by them in the same conversation
ToolsThey have all of them from day one — nothing to pick

Every change is shown to you as a card. Each time the new hire writes their own name, mission, or personality, a card appears in the chat showing exactly what changed, with a Revert button if you don’t like it.

Onboarding only ends when you confirm it

It doesn’t end on its own just because time passes. It ends only once the new hire finishes and you approve (Step 3 below) — until then, put off the interview and the new hire just waits.

No time to interview right now? Just leave it. If nothing moves forward for a long while (72+ hours), the progress card at the top of the chat switches to “Stalled” so it’s easy to notice — nothing to worry about, just say something to the new hire and it picks right back up. You can also reopen onboarding later from the agent’s settings.

While this interview is underway, the top of the chat shows a progress card with four stages — Joined → Deciding the work → Preparing the workspace → Assignment — marking how far along things are.

Step 3: Confirming ends the onboarding

When things are settled, the new hire shows you a summary and asks whether to confirm. Once you do:

  • The new-hire badge disappears and they become a regular employee
  • Working on their own initiative switches on
  • The onboarding-only ability to rewrite their own profile is taken back (name and mission are edited from Settings after this)

The new assistant introducing itself and proposing how it will work

Right after confirming, they’ll usually propose how they intend to work.

Morning Brief — Every morning at 8 I’ll go through the last 24 hours of AI-startup and semiconductor news, and give you the five that matter plus one line on the trend to watch today. Shall I set that up to run daily?

From here you do one of two things.

  • Happy with it → “Yes, do that.” → They register the schedule, and the briefing starts tomorrow morning.
  • Want changes → Just say so. e.g. “Make it 7:30 instead of 8”, “Drop the trend line”, “Include source links”

Employees can refine their own behavior instructions through chat, so you never need to open a settings screen to edit them.

Step 4: Try it once by direct instruction

If you want to preview the result before the auto-run kicks in, just ask the assistant on the spot.

Show me today's news briefing.

When it gets your request, the assistant runs tools like web search several times on its own to gather material, manages the work steps as a checklist, and then delivers an organized result. The chat shows which tools it used and how many times (e.g. web_search (13×)) alongside the progress.

A single request drives the assistant to run web_search 13 times, finish its work plan, and deliver an organized briefing

Keep it as-is if you like it, or refine it through chat:

  • “Lengthen the trend line to two lines”
  • “Drop the emojis”
  • “Group US-listed companies separately”

This is how the assistant works in two modes side by side: (a) autonomous run — on its own per the schedule — and (b) direct instruction — instantly when you ask.

Step 5: Adjust the schedule (optional)

Heartbeat — “look around on its own”, on by default

Your assistant has a heartbeat on by default. While the heartbeat is on, the assistant looks at the situation and decides on its own when to check in next — less often when there’s nothing to do, more often when something’s pending. So even without an exact schedule, the assistant wakes up and looks for work on its own. If you want it to only handle what you ask, turn the heartbeat off in the Schedules tab.

The heartbeat toggle at the top of the Schedules tab and the "Next heartbeat" time

Want it to run at a fixed time — Add a Schedule

If you want it to run at an exactly fixed time like every morning at 8 AM, add a schedule. If you accepted the assistant’s proposal in Step 3, it may already be registered. To change the time or the trigger message, just ask the assistant — or adjust it in the UI.

Ask the assistant:

Schedule this to run automatically every morning at 8 AM.

Set it up in the UI

Open https://upserve.app/agents/{agentId}/schedules (or the Schedules tab on the agent page) to register one. You never write a cron expression by hand.

  1. Click Add Schedule in the top right
  2. At the top of the form, pick Repeat (or Once for a single, one-time run)
  3. In the Prompt field, type the message the assistant should receive each time this schedule fires (e.g. Summarize today's news briefing) — this field is required before you can save.
  4. Choose a repeat rule — at a specific time (pick days, plus one time of day; Every day / Weekdays only shortcuts are provided) or on a fixed interval (pick one of 30 min12 hours, repeating all day)
  5. Click Add Schedule at the bottom of the form to register. From then on, every time this schedule fires, the message you typed in step 3 is sent to the assistant automatically, and it acts on its own behavior instructions — you don’t have to send a chat message yourself each time. To change what it does, edit the prompt, or just chat with the assistant: “update your behavior instructions like so.”

Tip — For a single fire at 8 AM daily, choose at a specific time, then set Every day + 08:00.

The timezone follows your profile setting (e.g. Asia/Seoul) automatically. The first push arrives tomorrow at 8 AM.

For the full breakdown of schedule types (repeat/once), limits, and safeguards, see Scheduling.

What happens every morning at 8 AM

⏰ Every day, 8 AM ┌──────────────────────────────────┐ │ Morning Brief — news assistant │ │ (running the mission it wrote) │ └──────────────────────────────────┘ │ ① Web search "AI startups (last 24h)" │ ② Web search "semiconductor market (last 24h)" ┌───────────────────────────────────┐ │ Pick top 5 + one-line summaries │ │ + one trend to watch │ └───────────────────────────────────┘ │ ③ Deliver to user 📱 Push notification ✉️ In-app chat message

Variations

The same pattern (hire → the new employee interviews you → they propose how to operate) works for employees like:

  • Competitor price assistant — “An assistant that keeps an eye on competitor price changes”
  • To-do reminder assistant — “An assistant that pokes me about pending items each morning”
  • Weekly report assistant — “An assistant that wraps up the week into a Friday report”

Advanced

Author the behavior instructions by hand

If you’d rather skip the full interview and write the behavior instructions yourself, this section is for you. It’s best suited to users who already craft instructions by hand in tools like Claude Code or Cursor.

After hiring, hand the new employee everything at once — “here’s the mission, write it up and confirm” — and they’ll do exactly that in a single turn, with no back-and-forth.

Once the agent is created, the items you can adjust directly in the agent’s Settings screen are Name · Mission (core purpose) · Description · Model · Enabled tools · Memory retention, and so on. Just write your intended purpose clearly in the Mission field, like this:

Organize AI startup and semiconductor market trends every morning so they can be read in 5 minutes. Focus on the few changes that can't be missed and what they mean for today.

UpServe has no separate box for writing a whole system prompt. You steer the detailed behavior (tone, format, rules) through chat, and the assistant folds it into its own instructions — just message the agent things like “update your behavior instructions like so” or “always attach source links”.

Tools: nothing to turn on — every agent is born with all of them, web search included. Schedule: Schedules tab → Add Schedule → under the repeat rule, choose at a specific time, then set Every day + 08:00 → register with Add Schedule at the bottom of the form. (Or leave the heartbeat on and let the assistant look around on its own.)

Installing a preset that needs customization

Some presets you install from the Market in the Community tab include items that need to be tailored to your setup — such as keywords or account connections. When you install one of these presets, a short interview screen appears first. Answer the questions and your personalized agent is created automatically. Once the interview is done, the UI takes you straight to the new agent’s chat screen.

For the full installation flow, see Installing Skills & Presets.

What does the manager do?

Your manager doesn’t hire. Hiring is the button in Step 1. The manager does two things:

  • Looks after the employees you’ve hired — ask how someone has been getting on, whether anything is stuck, how last week went, and they’ll check the actual run history before answering.
  • Answers questions about UpServe — pricing, limits, where a setting lives, how a feature behaves.

On the web they’re the button at the bottom right; on mobile, the manager icon in the top toolbar. Ask where something is and they’ll often point at the button on screen for you.

If you no longer want one, you can remove it from Settings → Manager (/settings/manager), and add a new one anytime from the same screen’s Add a manager button.

Removing it deletes your conversation history with the manager, and this cannot be undone. Adding a manager again later doesn’t bring the same one back — it starts fresh. AI employees you already hired are unaffected either way.

Next steps