Install AI Server on Windows

The Windows app is the quickest way to a working server. It installs the server, manages it, and gives you pages for models, API keys, governance and monitoring. On your own PC it is free.

Requirements

• Windows 10 version 2004 (build 19041) or later, or Windows 11, on x64 or Arm64.
• 8 GB of memory is enough to try small models; 16 GB or more is comfortable. A GPU with 8 GB or more of video memory makes chat, images and speech interactive for a team. See sizing.
• Disk space for models: a small chat model is about 1–2 GB, a 7–8B model about 5 GB, image models several GB each.

Install

1. Install AI Server from Microsoft Store.
2. Open it. The Server page shows whether the server is running, its address and who is connected.
3. Open Models and download a model. The app suggests a free starter model.

Every AI Suite app on the same PC now finds the server automatically and sends its AI work to it. Nothing else needs configuring for local use.

Check it works

curl http://127.0.0.1:11436/v1/models

A JSON list of models means the server is up. On this PC no API key is needed. Third-party tools on the same PC (scripts, chat UIs, editor plugins) work too; on Free they share a small allowance of 10 requests a day and 1 a minute per client. AI Suite apps are never limited.

Share it with your network

Serving other devices needs Pro Personal or Pro Commercial — see editions.

1. Subscribe in the app (Settings → Subscription), or enter a licence key bought from the Software Tailor Store.
2. Open API keys and create a key for each person, device or app. The key is shown once; the server keeps only a salted hash. Personal allows up to 5 keys; Commercial is unlimited.
3. On the Server page, under Server settings → Access, choose Every IPv4 network on this PC (or Every network, IPv4 and IPv6) and save.
4. Choose Install as service so the server keeps running when nobody is signed in and starts with Windows. Windows asks for administrator approval once.
5. Allow the port through Windows Firewall when the app offers it, or use Network diagnostics → Test connectivity.

Without a key, or without a paid licence, the server keeps serving this PC only and says why on the Server page — it never exposes an unauthenticated server.

Other devices connect to http://<server-ip>:11436/v1 with an API key. AI Suite apps and AI Client find the server on the local network and ask for the key.

Run as a service

The service runs as its own low-privilege account, NT SERVICE\AISuiteServer, with access to the AI Server data and model folders only. If Windows cannot start it under that account, the installer falls back to LocalSystem and records why in the install log.

• Start the server when AI Server opens (Server page) starts the in-app server at launch if you do not use the service.
• Changing the port, network access, HTTPS or rate limits needs the service to be reinstalled; the app says so and does it with one approval. Log level and governance changes need only a restart.
• After an app update the app offers to refresh the service so it runs the new version.

Where things are kept

Do not delete %USERPROFILE%\AISuite\v2: besides re-downloadable models it holds every AI Suite app's documents and your licence registration. To reclaim disk space, delete models or engines only.

Next steps

• Issue and scope keys: API keys
• Turn on HTTPS: TLS
• Point your code at it: API overview
• Something not working: troubleshooting

Questions