Frequently asked questions

Agent X is a virtual AI office that runs on your own computer. It places a team of Claude and Codex agents in your projects; you manage them from a 3D office with real terminals, a task board and reports.

Yes. Agent X runs on your computer and connects with your own Claude or Codex account. Model usage runs on your account; the Agent X subscription only covers the app and the office layout.

Windows 10 and 11 for now. You also need a Claude Code or Codex CLI account. macOS and Linux are not supported. To watch your office from a phone there is an Android app (APK) and a page that opens in the browser; details are on the Download page.

No, opening an account is free and does not ask for card details. You pay only when you buy a plan.

Three steps: open an account on the site, download and run the Windows installer from the Download page, and sign in to the app with your site account. Then you connect your own Claude or Codex account. The blog post "Kurulum ve ilk giriş" (in Turkish) walks through the steps.

The installer is not yet signed with a code-signing certificate, so Windows SmartScreen or your antivirus may show a warning. You can compare the SHA-256 value on the Download page with that of your own file to make sure it was not corrupted. Updates are additionally verified with the publisher's digital signature and checksum.

In Windows PowerShell run the Get-FileHash command with the SHA256 algorithm and compare the result with the SHA-256 on the Download page. If they match, the file is intact.

At sign-in, yes: the app is verified against your site account. After you sign in, if the connection drops you can keep working offline for up to 5 hours; after that you need to reconnect and verify again. The agents' own Claude or Codex connection needs internet separately.

When a new version is out the app shows an update notice. The downloaded installer is verified with the publisher's signature and its SHA-256 value. Your data and settings are kept during the update.

Depends on the plan: Lite 1, Pro 2, Max unlimited. You see and remove your devices on the My account page; you can remove at most 3 devices per month.

Lite 3, Pro 5, Max unlimited. This number is the plan limit.

Plans are monthly and sold in US dollars (USD). Each purchase extends your time by 1 month. When the time ends, your subscription ends; you can buy again. It does not renew by itself.

Details are on the Cancellation and refunds page. In short: the subscription does not renew by itself, so there is nothing to cancel; it ends when your time runs out.

On your computer. Agent X runs locally; your projects, agent conversations and files are not sent to this site's server. The app connects only for license and version checks. The agent's model (Claude or Codex) processes the data it reads while working under the terms of its own service.

No. Payment is taken on the secure page of the payment provider PayTR; card details are not stored on this site.

We set no cookie while you browse the site. Two exceptions: a language cookie that remembers your choice if you pick a language, and a required cookie that keeps your session open if you sign in. There are no advertising or tracking cookies. We count visits only as totals, without IP addresses or cookies. Details are on the Privacy policy page.

Steps that leave your computer (for example writing a public reply to a comment, or saving a Studio output) are asked first with an approval card. The card says what will be done, why, what it changes and its risk; you answer Yes or No.

The section that produces images and video on your own computer with your own graphics card: text to video, image to video, video to video, multi-clip, image generation, face swap and storyboard. Because generation runs locally, generating images and video does not spend tokens.

Windows and an NVIDIA graphics card. Each model also has its own video memory (VRAM), RAM and free disk requirement; for example FLUX.1 schnell needs at least 8 GB VRAM (12 GB recommended) and the large Wan 14B models at least 16 GB. If your card is not enough for a model, the list says so. A later version (0.1.34, not released yet) brings a slow "low memory mode" that lets weaker cards try.

The models are large: the approximate download sizes in the list are between 9 GB and 23 GB, and installing needs more free disk than that. Before installing you can set the model folder to another drive in Settings, and you can delete, move or open the folder of installed models.

Each model comes with its own license and the list shows it. For example FLUX.1 [dev] is for non-commercial use; the Qwen Research License of Qwen-Image based models is for research and commercial use needs legal review. Following the license is the user's responsibility.

Generation itself does not. If you choose Opus or Codex to translate your Turkish prompt into English, that translation spends tokens and your approval is asked first; local translation is free. A lead or researcher agent writing a prompt also spends tokens because the agent runs.

No. Generation for impersonation and political forgery is refused. Face swap is off by default and may be used only for your own face or for a person who gave written permission. Outputs get "AI generated" metadata.

The YouTube Manager role prepares a channel summary every morning at 09:00 (views, subscribers and comment changes), writes reply drafts for new comments and helps prepare a Studio output for upload. Steps such as sending, deleting and uploading are asked first with an approval card.

The Trendyol Manager role prepares a store summary (sales, orders, returns, stock) and a competitor price comparison every morning at 09:00 and writes reply drafts for customer questions. It is read-only: it does not change prices or stock; sending a reply needs an approval card.

Use the Suggestion / Complaint button at the top of the app: your message is tied to your account, you can attach a screenshot and you see the reply in the same place. Other ways are on the Contact page.

Write through the ways on the Contact page, stating the email address of your account. Which data we process and why is on the Privacy policy page.