Runs where your data has to live.
On-premises, in your own cloud, or both. Agents, workflows, guardrails and audit behave the same everywhere.
Pick one, or mix per agent.
Everything on servers you own.
Models, embeddings and the voice engine included. Nothing has to leave your building.
Your account, your region, your keys.
The same components in your own cloud account. You hold the account and the data.
Decide agent by agent.
Keep sensitive agents fully local and let the rest use cloud models.
Every component can be self-hosted.
There is no piece that only runs on our side.
Studio
Where teams build and operate agents and workflows.
Platform server
Teams, permissions and the connector library.
AI service
Agents, conversations and knowledge, with the separate audit database.
Voice worker
Speech recognition, the voice and turn detection, from local model files. No audio or transcripts written to disk.
Workflow microservices
The Java services compiled from your workflows, on machines you register.
A development shortcut cannot reach production.
The AI service refuses to start in production with a development encryption key, a development audit password, or no audit fingerprint key.
- A licence your servers check themselves: Ed25519-signed, no call home.
- Rather not run it yourselves? Talk to our team about hosted options.
Plan it around your network.
Tell us where your data has to stay. We will map each component to your servers, your cloud or both.