Outside providers that do not keep your data
Use only providers that agree not to store prompts or files or use them for training.
Private AI Infrastructure.
Give your team the AI models they need, with the privacy, budget, and control you choose. Setup is a one-time fee. An annual subscription covers operation, support, and maintenance.
- PrivateChoose your models and what you own
- ConfidentialChoose where and how AI runs
- ControlledSet access, usage, and team budgets
- PersonalizedAdd your tools, rules, and integrations
Run approved AI models in one private workspace.
You choose where your data goes and what is retained. Approve access and team budgets so AI spend stays focused on work.
Separate AI accounts→One private workspace
Provider controls data handling→You choose where data goes
Open-ended usage→You approve access and budgets
300+ models. One private workspace.
Your setup determines which models are available. Approved external model providers retain none of your data.
GPT · OpenAI
Hunyuan · Tencent
GLM · Z.ai
DeepSeek
MiMo · Xiaomi
Gemini · Google
Solar · Upstage
Kimi · Moonshot AI
MiniMax
Claude · Anthropic
Ling · inclusionAI
Grok · xAI
Qwen
Mistral
Llama · Meta
Nemotron · NVIDIAThree ways to run AI privately.
Astraed manages and operates all three options. You control workspace privacy; approved external model providers retain zero data.
Private servers in Europe
Run models on dedicated servers in the EU, Switzerland, Norway, or Iceland.
At your site or offline
Run models on your hardware, with no outside network if needed.
Choose the provider and country.
ScalewayFrance · EU
OVHcloudFrance · EU
HetznerGermany · Finland
InfomaniakSwitzerland
STACKITGermany · Austria
ExoscaleSwitzerland · EU
UpCloudFinland · Norway · EU
Bulk InfrastructureNorway
VerneIceland
How we get your private AI infrastructure running.
From a 1–2 week pilot to infrastructure managed and operated by Astraed.
- 01
Assess
Assess employee needs, privacy, and optional development.
- 02
Architect
We configure models, hosting, access, and budgets.
- 03
Pilot
Use the setup for 1–2 weeks, then decide whether to continue.
- 04
Operate
We operate, support, and maintain your infrastructure.
Optional services for your team.
Custom models, fine-tuning, tools, and integrations are optional and quoted separately.
Models tuned to your work
Match your terminology and standards.
Task-specific tools
Turn repeatable methods into reusable tools.
Integrations
Connect approved information and systems.
Ways of working
Set clear steps and human review.
Different work. Different controls.
Choose who can use AI, where it runs, and what it can access.
Frequently asked questions.
How does Astraed keep our data private?
Every client gets a dedicated server, never shared with another client. Astraed manages and operates it. You control workspace access and retention; approved external model providers retain zero data.
Can every team use its approved models in one place?
Yes. Combine approved external and privately hosted models in one workspace. Custom models and fine-tuning are optional. Offline systems use local models only.
Can our data and models stay in Europe?
Yes. We can use infrastructure in the EU, Switzerland, Norway, or Iceland.
Can Astraed run entirely at our site or offline?
Yes. Astraed manages and operates on-premise, offline, and air-gapped deployments. Offline and air-gapped systems run local models without sending data to external model providers.
Are prompts stored or used to train models?
Approved external model providers retain zero data and do not use it for training. You control workspace privacy, including access, conversation history, logs, and retention.
Can we control models, access, and spending by team?
Yes. Set team budgets and review usage by team and model. Budget-limit behavior and the costs covered are defined for your selected setup.
Can Astraed connect to our existing data sources and tools?
Integrations are optional. We scope connections to your approved files, databases, identity systems, and tools around your access rules.
Does our team need technical expertise?
No. We design, deploy, and operate the infrastructure; employees use one workspace.
Can we change models or hosting later?
Yes. Approved models, limits, and deployment can change with your needs.
How does the pilot work?
We assess employee needs, privacy requirements, and any optional development. Your team then uses the setup for 1–2 weeks and decides whether to continue. Pilot scope and any fee are agreed in advance.
What does the annual subscription include?
Ongoing operation, support, and maintenance, including on-premise installations. Setup is charged separately as a one-time fee. Your teams use the workspace; we run the system.
How is the annual subscription priced?
Pricing depends on infrastructure, privacy requirements, and employee bands. External model usage, optional custom development, and hardware leases are billed separately.
How is model usage billed?
External model usage (inference) is billed after each month at provider cost, with no markup. We do not charge for inference from local models running on your premises.
Can we lease the hardware?
Yes. On-premise hardware can be leased under a separate contract and bill from the annual subscription. Astraed still manages and operates the setup.
Plan your private AI.
Tell us your team size and privacy requirements. We will scope your pilot and agree any fee before it starts.