Data location and retention
Choose where business data, files, memory, logs, and backups are stored and how long they are retained.
Deployment options
Choose the operating model that fits your requirements
Balance infrastructure control, operational responsibility, and speed without changing the way agents work.
| Capability | Customer-hosted | Private managed | UnaGo Cloud |
|---|---|---|---|
| Infrastructure ownership | ✓ Your team | Shared operating model | UnaGo |
| Data boundary | ✓ Customer-defined | Dedicated environment | UnaGo-managed environment |
| Model options | ✓ Self-hosted, private, or approved external | Private or approved external | Supported cloud models |
| Internal systems | ✓ Private network access | Private connectivity | Secure integrations |
| Operational responsibility | ✓ Highest control and responsibility | Balanced | Lowest customer overhead |
Data privacy
Know exactly where data goes
Self-hosting gives your organisation control over the platform boundary. It does not automatically mean that no data leaves that boundary: configured external models, integrations, telemetry, or update services may still receive data. Full containment requires every dependency and egress path to be hosted or restricted inside the approved environment.
- Keep storage, memory, files, logs, and credentials under customer-defined policies.
- Control network egress and approve which models and systems may receive data.
- Record agent activity, tool use, approvals, and outcomes for review.
- Define backup, recovery, retention, and deletion around internal requirements.
Private execution
Connect agents to the systems your business depends on
Let autonomous teams operate across private and existing software while access remains limited, observable, and approval-controlled.
Internal APIs
Connect private services and custom applications through approved interfaces.
Governed data and knowledge
Use internal documents, databases, and approved datasets without copying them into disconnected tools.
Secure browser execution
Run browser work in isolated environments with controlled destinations, files, and credentials.
Secure cloud computer
Operate legacy or browser-only software where an API is unavailable.
Company-scoped secrets
Keep model and integration credentials inside the customer security boundary.
Approval-controlled actions
Pause before submissions, transactions, publishing, or external communications.
FAQ
Enterprise deployment questions
Clear answers about containment, models, operations, and internal connectivity.
Can UnaGo run without public internet access?
A fully offline or air-gapped configuration requires models, integrations, telemetry, licensing, and update paths to operate inside the approved boundary. Confirm these dependencies during deployment design.
Which self-hosted models can we use?
UnaGo is designed around configurable providers and private or OpenAI-compatible endpoints. The chosen model and inference server should be validated for tool use, context size, quality, latency, and hardware requirements before production.
Does any data leave our environment?
Only the services you configure should receive data. External models or integrations may receive selected request content, so model routing and network egress must follow your data policy.
Can departments use different models and tools?
Yes. Workspaces can separate department data, agents, memory, files, integrations, and permissions, with central governance above them.
Can agents connect to systems that are not publicly accessible?
A customer-controlled deployment can connect to approved internal APIs, databases, repositories, and browser-based systems within the permitted network boundary.
Who operates the self-hosted platform?
The operating model is agreed for each deployment. It should define responsibility for infrastructure, scaling, monitoring, backups, upgrades, models, integrations, security response, and support.