Cloud
Run M3 Forge as a managed service operated by Marie AI. This is the SaaS model: we operate the control plane, the runtime, and the backing infrastructure.
What Cloud means
In the Cloud model:
- Marie AI runs M3 Forge in our cloud
- Marie AI runs the Marie AI data plane in our cloud
- Marie AI manages upgrades, scaling, observability, and platform operations
- customers use the product without installing Kubernetes components
Responsibilities
| Layer | Who operates it | Notes |
|---|---|---|
| M3 Forge UI and APIs | Marie AI | Multi-tenant control plane |
| Marie AI runtime and gateways | Marie AI | Shared managed runtime |
| PostgreSQL, object storage, ClickHouse | Marie AI | Operated as managed platform dependencies |
| Customer workflows, prompts, org configuration | Customer | Managed through the product |
Architecture
Best for
- teams that want the fastest path to adoption
- customers that do not want to run Kubernetes for the platform
- internal Marie AI-managed environments
- organizations that want managed upgrades and operations
Security and network posture
The exact production topology depends on region and hosting environment, but the Cloud model is expected to provide:
- managed ingress and TLS
- managed PostgreSQL and analytics infrastructure
- centralized platform observability
- region-aware control-plane and runtime placement
- explicit egress and allowlisting guidance for enterprise deployments
Reference model
LangSmith documents Cloud as a fully managed model where the vendor hosts both the UI/API and the runtime workloads, and also documents regional hosting and private connectivity options for enterprise users. We are following the same hosting-family split for M3 Forge. Source:
Current status
The Kubernetes packaging work for the Cloud-capable control plane is underway. The standalone m3forge chart exists today as the first slice of that work, but the full managed-cloud operating model is still being formalized.