---
title: "AI employees that run locally | AgentVUE"
description: "Hire AI employees that run on your machine: they execute recurring digital work with your models and your keys. Definition, comparison table and straight answers. From €49/month."
canonical: "https://agentvue.ai/ai-employees/"
language: "en"
---
AI employees

# AI employees that run locally.

An AI employee holds the job: it plans the work, runs it on your machine and comes back with the result. You keep the decisions.

[Get AgentVUE→](/pricing/)[Watch the 90-second tour](/product/tour/)

-   Runs on Windows & Linux
-   Your models, your keys
-   No account. No telemetry.
-   From €49/month

AgentVUE DesktopLocal

HomeAgentCodeAgentWorkShortVUECyberBusiness Kit

AgentWork / This week**Weekly exposure review**

01Inventory the machine✓

02Match published advisories✓

03Prepare the fix

Cyber engineerCyber analystSocial manager

Awaiting your approval · report on disk

Local**0 bytes sent**

Approval**Waiting for you**

What it is

## A job, not a conversation.

A chatbot answers and moves on. An AI employee keeps the task list, the context, the last run and what is still pending.

ChatbotAI employee

Same request.

What is exposed on this machine?

I can’t see your machine. Paste the details here and I’ll answer.

The conversation ends when you close the tab.

**Weekly exposure review**On your machine

-   01Inventory the machine
-   02Match the published advisories
-   03Prepare the fix
-   04Wait for your approval

The report stays on your machine. You decide what changes.

### Holds the job

The task list, the context and the last run stay together.

### Runs the work

Code, media, browser and Git on your machine.

### You decide

Sensitive actions wait for your approval.

### Stays local

Your files, models and keys stay where they are.

It keeps the task list, the context and the last run. You delegate an outcome, not a prompt.

A chatbot answers and forgets when you leave. An AI employee comes back with the work done, on the machine where the files already live.

[See how it compares](/compare/)

Compared

## AI employee vs chatbot vs cloud AI employee

Three products get called “AI” in the market. They do not belong in the same column.

AgentVUE compared with a chatbot and a cloud AI employee
| Aspect | Chatbot | Cloud AI employee | AgentVUE AI employee |
| --- | --- | --- | --- |
| Where it runs | A browser tab on the vendor's site. | The vendor's servers. | Your Windows or Linux machine. |
| What it does | Answers questions inside a conversation. | Runs tasks through the vendor's tools and connectors. | Executes recurring digital work on your machine and repositories. |
| Models and API keys | The vendor's model, no choice. | Vendor-hosted, or your key routed through their platform. | Your models, your keys, your endpoints. |
| Your data | Held by the vendor. | Held by the vendor, in their account. | Stays on your machine unless you export it. |
| If the vendor disappears | The conversation ends. Little is lost. | The employee stops the day the vendor switches the service off. | It keeps working: runtime, licence and data live on your machine. |
| Human approval | Not applicable. | Depends on the platform. | You approve what matters and set the autonomy level. |

[Compare AgentVUE with the field](/compare/)

How it works

## How a run goes.

Four steps. All on your machine.

Task loopPlay the loop

01**Inventory first**02**Prioritised work**03**Approval where it matters**04**The report stays with you**

### Inventory first

The workstation maps your machine and your repositories: installed packages, services, projects, versions. The agent works from facts on your disk, not from a guess.

### Prioritised work

What matters most is handled first — the exposed version before the nice-to-have, the client deliverable before the internal cleanup. Recurring work can become scheduled tasks: every Monday, the brief is prepared and waiting.

### Approval where it matters

A fix is applied with your approval. A publish runs through the platform's own web interface, on your accounts — by the agent itself, with a review gate whenever you want one. You set the autonomy level once; you can change it whenever you want.

### The report stays with you

Each run ends with a report on your machine — what was checked, what changed, what is pending — ready to hand to a client or an insurer questionnaire.

Who decides

## You approve what changes something.

The agent can plan, run and check the work. A fix on your machine, a commit, a publish on your account: those wait for you when you want them to.

The agent runs the checks, prepares the content and keeps the task moving. You approve the actions that change something: a fix on your machine, a commit, a publish on your account.

Set the autonomy level once. Change it whenever you want. The report stays on your machine.

Autonomy

### Set the gate where you want it.

Keep the approval step on every change, or let routine work run. You can change it at any time.

TaskYour approvalAction

Ask me first

The action waits for your approval. Nothing changes until you say so.

The work

## It already handles the week.

The same workstation covers code, operations, security and content.

[All use cases](/use-cases/)

### Run IT operations

Level-1 triage, machine inventory, services and open ports, fixes proposed then applied with your approval.

A fleet you can actually see, fixes that stick.

### Customer support

Answers grounded in your knowledge base, escalation of what needs a human, tickets updated automatically.

Less copy-paste, answers that match your docs.

### DevOps and releases

CI failure triage, release notes, deployment checklists and pipeline documentation.

A release train that keeps moving.

### Compliance and audit prep

Control mapping, policy upkeep and the audit-prep file built through the year.

The file is ready before the audit is announced.

Roles

## Pick the job you want off your plate.

Three roles ship today. Same licence, same machine.

Cybersecurity engineerCybersecurity analystSocial media manager

AgentWork · CybersecurityEngineer

This run**Weekly exposure review**

-   Package inventory**142 packages**
-   Advisory match**3 findings**
-   Fix prepared**PHP 8.4.27**
-   Awaiting approval**1 fix**

InventoryAdvisoriesFixPost-check

Critical**1**High**2**Local only**0 B**

Snapshot taken. Waiting for you.

AI employee

### Cybersecurity engineer

Designs and applies the fixes: machine inventory, advisory matching, the patch applied with your approval, then the post-check.

-   Prioritised fixes for exposed services
-   Verified patches applied with your approval
-   A dated report

[Open the role page→](/ai-employees/cybersecurity-engineer/)

Cyber analyst · ReportRead-only

This run**What is exposed**

-   Critical**PHP 8.4**
-   High**Node.js**
-   Medium**2 packages**
-   Report**Dated · local**

InventoryMatchExplainHand over

Findings**5**Explained**5**Applied**0**

Read-only. Nothing is applied.

AI employee

### Cybersecurity analyst

Reads the exposure and explains it: what is vulnerable, which version fixes it, what to treat first.

-   An inventory you can actually use
-   Prioritised findings
-   Advisory reading in plain language

[Open the role page→](/ai-employees/cybersecurity-analyst/)

ShortVUE · This weekSocial

This week**Weekly content**

-   Brief**Ready**
-   Assets**4 produced**
-   Variants**3 channels**
-   Publish**Review gate**

ScriptVoiceSubtitlesVariants

**This week**9:16

Format**9:16**Channels**3**Local**Yes**

Ready to publish. Waiting for you.

AI employee

### Social media manager

Runs the content pipeline: shorts, images and per-channel variants, then publishing through the platforms' web interfaces — by the agent itself, on your accounts, on a schedule you define.

-   A recurring content cadence
-   The short-video pipeline
-   Per-channel variants

[Open the role page→](/ai-employees/social-media-manager/)

Start here

## Hire your first AI employee.

One licence. One machine. Start with the job that costs you the most time.

[Get AgentVUE→](/pricing/)[Compare AgentVUE](/compare/)

Built with AgentVUE

## AgentVUE builds AgentVUE.

The editor, the websites and the Business Kit are built and maintained by the same agents, on the same workstation.

-   Windows
-   Linux
-   AgentCode
-   AgentWork

[See the showcase](/showcase/)

![AgentCode showing a GitLab work item to build the one-page service site, the AgentVUE Suite sidebar and an open positioning.md draft.](/images/product/agentcode.webp)

**AgentVUE builds AgentVUE**AgentCode building the AgentVUE site from a GitLab work item, with the Business Kit in the sidebar.

Keep reading

## Where AI employees already earn their keep

Two specialisations are live today, on the same licence. Both run on your machine, start to finish.

[The cybersecurity track](/cybersecurity/)[The marketing track](/marketing/)[Plans and pricing](/pricing/)[AgentVUE vs the field — every comparison](/compare/)

Questions

## AI employee questions, answered

The licence, the data, the approvals and the price.

What is an AI employee, in one sentence?

Software that holds a recurring job on your behalf: it plans the work, executes it on your machine and reports back, while you approve what matters.

How is an AI employee different from a chatbot?

A chatbot answers questions in a conversation and moves on. An AI employee keeps the job: the task list, the context, the previous runs. You delegate an outcome, not a prompt.

How is a local AI employee different from a cloud one?

A cloud AI employee runs on the vendor's servers, with the vendor's tools, and your data sits in their account. A local AI employee runs on your machine, with your models and your keys, and the data stays on your disk unless you export it.

What if my AI employee's vendor shuts down?

It already happened once. Vellum announced that its Personal Assistant Platform shuts down on 2 November 2026: cloud-hosted assistants cease to function, logins are disabled and the cloud data is permanently deleted. Vellum itself tells users to self-host. A local AI employee keeps working, because the runtime, the licence and the data live on your machine. Every cloud vendor can be the next one — that is a structural risk of the category, not a one-off.

[Vellum — Assistant Platform shutdown notice](https://www.vellum.ai/vellum-assistant-platform-shutdown)

Does an AI employee replace a person?

It does the recurring work; the human approves what matters. Treat it as capacity, not headcount: the checks run, the content is prepared, the updates are applied with your approval — and the judgement stays yours.

What does an AgentVUE AI employee need to run?

A Windows or Linux machine, an AgentVUE Desktop licence and an AI endpoint you configure — your models, your keys. There is no account, no activation server and no telemetry.

Can it work on my company's repositories?

Yes. The agent works on the machine where the repositories live, with your Git configuration — GitHub and GitLab are supported by the task manager. You keep control over what the agent may touch.

What does one cost?

One licence covers the whole workstation: Pro €49, Max €99, Ultra €199 per month billed yearly, or €33/€66/€123 per month with three years prepaid and one year free. The details are on the pricing page.

Is my data sent anywhere?

No. The software creates no account and sends no telemetry. Your files, findings and reports stay on the machine unless you explicitly export them.

Can an AI employee publish on social media?

It prepares the content locally — scripts, cuts, subtitles, per-channel variants — then publishes it itself through the platform's own web interface, on your accounts, never through a platform API. Recurring tasks keep the cadence on schedule, and you can keep a review gate before anything goes out.

Next step

## Put the first job to work.

One licence, one machine. Install, configure your AI endpoint, and put the first recurring task to work this week.

[Get AgentVUE→](/pricing/)[Talk to the founder](/consultancy/)
