AgentFlow runs where your data and your governance say it has to run. The same components deploy everywhere; identity, storage, ingress, telemetry and model-provider settings change by environment.
SaaS
BYOC
In-VPC
On-prem
Other clouds
e6data operates the deployment. You connect your data sources and use the product with the lowest operational burden.
Bring your own cloud. The AgentFlow control plane runs in your cloud account. It can use your identity, networking, storage, telemetry and model-provider configuration.
AgentFlow runs inside your VPC or VNet, with ingress and egress controlled by you. Updates are delivered through the deployment process agreed for that environment.
AgentFlow runs on private Kubernetes infrastructure, including on-premises and hybrid environments. S3-compatible storage such as MinIO can provide the object store.
The architecture is portable across supported Kubernetes environments, including sovereign and non-hyperscaler clouds. Storage, identity, ingress and managed-service choices vary by platform.
What every deployment needs
- Kubernetes, for the API gateway, worker pods, sandbox pods and cron triggers (a local Compose profile runs without Kubernetes, but without cron triggers)
- Durable storage: S3-compatible storage, Azure Blob Storage, or PostgreSQL for product storage
- Redis, for live coordination
- At least one model provider
- An e6data cluster
See Architecture for how these fit together, and Before you begin for the setup checklist. Last modified on October 1, 2026
