Network & Data Flow

Superatom is deployed entirely within customer infrastructure. All external communication is outbound HTTPS only. No inbound ports need to be opened.

​

Outbound Traffic

The platform makes two categories of outbound HTTPS calls: Both destinations can be routed through the organization’s existing forward proxy or firewall.

​

What Never Leaves the Network

The following data categories are processed entirely within the customer’s infrastructure and never transmitted externally:

• Raw data rows or individual records
• PII or sensitive business data
• Database credentials
• Query results
• File contents
• User identities or conversation content

​

No Inbound Connections Required

Superatom requires no inbound ports to be opened. All external communication is initiated outbound by the platform. There are no webhooks, callbacks, or remote access channels.

​

Air-Gapped Deployment

For environments that cannot make any external calls, Superatom supports deployment with a self-hosted LLM running within the customer’s network. In this configuration:

• Open-source or licensed models run on the customer’s GPU infrastructure
• The Main Agent calls the LLM over the internal network
• No external API calls are made
• All telemetry is disabled or routed internally

​

What the LLM Receives

The LLM is a stateless inference service. Each request is constructed from scratch, and no state persists between calls.

​

Sent to the LLM

​

Never Sent to the LLM

​

No Training on Customer Data

Customer data, queries, and knowledge nodes are never used to train or fine-tune models. The LLM provider receives stateless inference requests only. This applies to both cloud-hosted and self-hosted LLM configurations.