Private AI. Always on. Under your control.
Three pillars, one platform. Data sovereignty, operational continuity and total governance: an AI that is truly yours must guarantee all three, without compromise. isolAIted was built on this certainty.
We believe artificial intelligence should work for you, not above you.
Our mission is to remove as many barriers as possible to rapid and effective AI adoption, without compromising privacy, control or economic sustainability.
Privacy by architecture
A contractual promise isn't enough. When a system blocks outbound connections, your data physically cannot leave. It's not a policy, it's an architectural property.
Zero marginal cost
Your hardware runs 24/7 without paying for every single query. No tokens, no rate limits, no surprise invoices. The investment is in hardware, once.
AI accessible to everyone
No CLI, no code, no technical expertise required to get started. We want anyone, from a solo professional to an enterprise team, to adopt AI in minutes, not months.
AI that empowers people
Reducing AI costs frees budget to keep investing in people. We want companies to use AI to amplify human talent, not replace it.
Geopolitical independence
A foreign cloud provider can change pricing, policies, or jurisdiction overnight. Your local infrastructure should not have this vulnerability. Your AI is yours, full stop.
No vendor lock-in
Open-source models, your infrastructure, your control. You can always choose the model, swap it, migrate. The platform creates no dependency: it's built for your freedom.
Security & Reliability
Continuous review by the open-source community means bugs and vulnerabilities are found and fixed faster than in closed systems.
Offline operation
The system can run without an internet connection — useful in isolated environments, edge computing, or with very low-latency requirements.
Governance & Total Control
Define who uses AI, how, and with what data. Granular roles and permissions, configurable isolation levels, versionable agents, and native audit trails. AI adapts to your organisation, not the other way around.
Not another AI interface. A paradigm shift.
isolAIted is not yet another wrapper around an inference engine that manages open models. It is the first platform to bring advanced agentic engineering out of specialist environments, making it accessible to everyone.
Intelligent setup
The first barrier to on-premise AI adoption isn't cost, it's technical complexity. isolAIted automatically detects available hardware, briefly collects your requirements on some fundamental aspects of the system, and makes anyone productive in 10 minutes, without ever opening a terminal.
AI Agents as business assets
In isolAIted an agent is not a simple system prompt. It's a fully configured entity with long-term memory that learns by applying itself to different tasks over time — a true "digital employee" that accumulates experience, knows the company context, and is available to the organisation, not just to a single project.
Collaborative AI Engineering
Agents, Workflows, Prompts and configurations in isolAIted are versionable and portable entities, not data locked in local databases impossible to migrate. Treating agents like code means applying the "Infrastructure as Code" approach to AI: a team collaborates on agents the same way it collaborates on software.
Three Interfaces. One platform.
Fast chat, simplified AI Agent orchestration in Workflows, agents working in full autonomy: three distinct operational levels that coexist in isolAIted's single graphical interface. Not a development framework, not a tool just for software developers, a ready to use platform, easy to enhance.
The intelligence you build today is worth something tomorrow and to everyone.
Everything you build in isolAIted — agents, workflows, prompts, automations, configurations — doesn't belong to a project, but to your organisation. isol Registry is where these assets accumulate, get versioned and become shared knowledge: the team always uses the latest approved version, best practices spread instead of staying on one person's laptop. AI stops being an individual tool and becomes a collective competitive advantage, one that grows over time.
Compliance by Design, not by Policy.
With isolAIted, regulatory compliance is not just a contractual promise: it is a structural property of the architecture. No data leaves, every operation is traceable, and the Deployer is guaranteed full control over personal data and the risk management of the system.
GDPR Compliant
isolAIted complies with the Privacy by Design and Privacy by Default principles under Art. 25 GDPR. The on-premise architecture ensures that the Deployer's personal data never leaves their own infrastructure, while dedicated tools allow users to exercise their GDPR rights.
- ✓ GDPR roles defined by architecture: the Deployer is the sole Data Controller; the Provider never accesses processed data. No DPA required with the Provider or foreign cloud providers.
- ✓ Privacy by Design and by Default (Art. 25): isolAIted is built from the ground up to protect privacy, integrating anonymisation filters where possible, with the highest level of privacy as the default setting.
- ✓ Data subject rights management (Arts. 15-17): the system enables users to exercise their GDPR rights, including the right of access, rectification, erasure (right to be forgotten), and objection to processing.
- ✓ Data minimisation and purpose limitation (Art. 5): the system processes only the personal data strictly necessary for the intended purpose. The Audit Log records usage metadata, not the content of interactions.
AI Act Ready
EU Regulation 2024/1689 introduces transparency, traceability and human oversight obligations scaled by risk level. With isolAIted every chat, workflow and automation is assigned a specific risk class: concrete technical measures follow automatically from that classification at runtime.
- ✓ Differentiated technical measures by risk class: the runtime automatically adapts controls and constraints based on the regulatory category of the model or workflow in use
- ✓ Transparency obligations (Art. 50 AI Act): notification of AI system interaction and marking of AI-generated or AI-manipulated content
- ✓ Voluntary codes of conduct (Art. 95 AI Act): native application of high-risk system requirements to limited-risk systems
- ✓ Italian Law 132/2025 (national AI Act transposition): anthropocentric principle guaranteed by native Human-in-the-Loop — no automated decision without human oversight
Open Weights at the core.
isolAIted works with hundreds of open weights models: Qwen, Llama, Nemotron, Gemma, Kimi, DeepSeek, Mistral and many more. No dependency on a single model provider. The choice is yours.
We believe in accessibility as a prerequisite for trust. You can choose the model, configure every parameter, inspect system behaviour. No black boxes.
- ✓ Compatibility with hundreds of open weights LLMs
- ✓ Integrated Model Library and Custom Model Editor
- ✓ Configurable parameters per individual model
- ✓ Integrated benchmark to test your own custom models
Your hardware. Your local network. Your speed.
High availability. Zero dependencies.
When your AI runs on your local network, every query travels centimetres, not across continents. The response arrives in milliseconds, not seconds. And it never stops.
No provider outages. No service degradation. No maintenance windows scheduled by someone else. Your infrastructure, your control, your internal SLAs.
Frequently asked questions
Not in absolute terms, and that's an honest answer. The system administrator can always configure the so-called "Isolation Level", which allows controlled exceptions: for example, authorising a specific agent to access an external API to retrieve updated data. It's a deliberate governance mechanism, not a flaw: every exception is intentional, logged and revocable. In the default configuration — maximum isolation — no data leaves. Control always stays with whoever administers the system.
Not at all. Every element that describes an agent in isolAIted also exists in the traditional approach — they just weren't separated. In fact, you'll quickly discover how much more efficient it is to keep agent definitions separate from individual projects: agents become reusable resources, not one-off configurations. For those who can't let go of the old schema, it's even possible — though not recommended — to simulate a Project-centric paradigm in isolAIted by creating a new agent for each project. But those who try it rarely go back.
Absolutely not. All open weights models are physically downloaded by isolAIted onto the user's PC or local server. Once installed, the model resides entirely within your infrastructure — no remote server is involved during use. The maker can stop distributing the model tomorrow: your instance will continue working exactly as before, with no interruptions and no dependency on anyone.
Ready to get started?
Download isolAIted for free and launch your private AI system in minutes.