Skip to main content
SynapsAI Cloud is one platform with three deployment modes. The API, Python SDK, agents, and knowledge bases stay aligned; what changes is who operates the GPUs and where customer content resides.

​
At a glance

Product overview for buyers: Sovereign and Security & data residency.

​
Hosted SynapsAI Cloud

  1. Launch a model from Hugging Face, an artifact, or a custom image.
  2. Call https://api.synapsai.cloud/v1 with your API key.
  3. Optionally attach knowledge bases and agents.
Inference traffic uses HTTPS. Workloads run on SynapsAI-operated GPU capacity with process-level isolation. Store requests is opt-in per endpoint. Details: Data, Security policy, Subprocessors.

​
BYOC (bring your own cloud)

Run the same platform inside your AWS, Azure, or GCP account. You keep cloud billing relationships and network boundaries; SynapsAI provides the runtime that builds, schedules, and serves models.
  • Same OpenAI-compatible API shape as hosted Cloud
  • Point the Python SDK at your Sovereign / BYOC base URL via SYNAPSAI_API_BASE
  • Scope residency and logging to policies you already enforce in that account
Contact support or synapsai.cloud/contact-us to scope a BYOC rollout.

​
On-prem (Sovereign)

Install on GPU servers you own—data center or colo. Models, prompts, embeddings, knowledge-base documents, and opt-in request logs stay on infrastructure you control.
  • Full platform parity with Cloud (agents, KBs, replay, RBAC, analytics)
  • Hybrid option: pin regulated tenants on-prem while bursting to Cloud
  • Air-gapped and tightly firewalled installs are scoped during enterprise discovery
See the marketing architecture and FAQ: synapsai.cloud/sovereign.

​
Choosing a mode

1

Prototype on hosted Cloud

Validate models, agents, and connectors without provisioning GPUs.
2

Decide residency requirements

If prompts or documents cannot leave your network, plan BYOC or on-prem.
3

Keep clients unchanged

Switch SYNAPSAI_API_BASE (and keys). See Migrate from OpenAI.