The answers link to the page with the details.
Questions
What is AI Server? +
A private AI server: it runs chat, embeddings, image, speech and vision models on your hardware and serves them through one OpenAI-compatible API to your staff, your apps and your code. See the overview.
Does my data leave my network? +
Not on the local-model path: prompts, documents and answers go between your clients and your server. Licence activation sends the licence key and random device ids, never content; cloud models are used only if an operator configures them. See security.
Is it free? +
On one computer, yes: the Free edition serves AI Suite apps without limits and other tools with a small daily allowance. Serving a network needs Pro Personal (US$9.99 a month) or Pro Commercial (US$49.99 a month per server). See editions and pricing.
Which licence does a business need? +
Pro Commercial, one per running server. An AI Gateway in front of several servers needs no extra licence; only the servers (workers) count.
What hardware do I need? +
A recent PC to try it; one workstation with an 8–12 GB GPU serves a team of dozens for typical office use. See sizing.
Does it run on Linux or macOS? +
Linux runs the Docker images and the Kubernetes chart. There is no macOS server app yet; macOS clients use a server on the network. See compatibility.
Will my OpenAI-based code work? +
Yes, by changing the base URL to your server's …/v1 and using an AI Server API key. Some newer OpenAI endpoints are not available yet. See the API overview.
Which models can I run? +
Open models from the catalogue — chat, embedding, image, speech and vision models — and any GGUF model from Hugging Face. See engines and models.
Can it work without internet? +
Answering requests does not need the internet. Model downloads and licence renewal do; a licensed server keeps serving through a licence-service outage of up to 7 + 7 days. See licence activation.
Do I need AI Server to use AI Suite? +
No. AI Suite apps have a local engine built in. AI Server lets many people and apps share one set of models.
What is the AI Gateway? +
A mode of AI Server (Pro Commercial) that puts one address in front of several servers, with failover, model-aware routing and upgrades without downtime. See scaling and HA.
How do I give someone access? +
Issue them an API key on the server and send it to them securely. You can limit it to certain models, endpoints and a rate, and revoke it any time. See API keys.
Can I see who uses it? +
Yes: usage per key, model and app, and a content-free audit log. Prompt and answer text are never recorded. See monitoring.
How is it updated? +
The Windows app through Microsoft Store; containers by pulling a new image. Engines update themselves between restarts. See upgrades.
Where do I get help? +
See support.