Your servers or private cloud
Run the runtime in infrastructure your organization controls. Your IT team plans networking, storage, backups, updates and model access.
Explore self-hosted →Security / Deployment
Your infrastructure.
Your AI boundaries.
Choose where agents run, what they can access and how work is reviewed. Account for models, connected tools and logs when defining your data boundary.
Your deployment boundary
LibraOSKnowledge · context · governance
Libra DeskLocal models
Connections you approve
External models & services
Conceptual example. An air-gapped configuration keeps dependencies local and disables external connections.
DEPLOY
Choose where it runs.
Choose where it runs.
Know who operates it.
Start with where your data can be processed and who will maintain the environment. LibraOS Cloud is currently a private beta.
Air-gapped environment
Use local models and dependencies. Keep connectors within the network, disable external calls and plan how verified updates enter the isolated environment.
Read deployment docs →A hosted deployment
Request access through the LibraOS team. Confirm region, model providers, retention and support arrangements for your beta deployment.
Request beta access →Follow the data
- Documents & knowledge
- Identify the storage and index used by your deployment. Assign knowledge access to the agents that need it.
- Model requests
- Local models process requests locally. An external model endpoint receives the prompts and context sent to it.
- Connected tools
- Each connected service has its own data boundary. Check what an action sends and which permissions it uses.
- Logs & backups
- Include audit records, monitoring destinations and backups in the deployment’s access and retention plan.
Assign the responsibilities
- IT & deployment owner
- Maintains a self-hosted installation and approves network, model and integration connections.
- Workspace administrator
- Manages people, agents and access to company knowledge.
- Business owner & reviewer
- Defines the task and reviews proposed actions within their authority. Checks the execution result after approval.
- LibraOS team
- Provides the runtime and deployment guidance. Agree support and hosted operating responsibilities for your deployment.
Self-hosting controls where the runtime runs. Review external model endpoints, tools and monitoring connections separately when defining your data boundary.
Governance / In everyday work
Make control visible.
Governance belongs in the task: which knowledge is available, who can act, and what happened after a decision.
Knowledge access
Assign with intent
Control the knowledge an agent can use. Keep organization document management distinct from agent access.
Human review
Decide with context
Inspect the proposed action and its recipient or target. Authorized reviewers decide where the workflow requires approval.
Audit & outcomes
Check the result
Keep requests, decisions and outcomes connected. Investigate failed or unknown results before taking another action.
Start with your work
Bring your deployment requirements.
Review infrastructure, model connections, access rules and operating responsibilities with our engineering team.