SDK vs SaaS: Why We Chose Both

Neurostack operates two products with fundamentally different deployment models: an on-premise SDK for organizational intelligence and a cloud SaaS for marketing analysis. This wasn't a compromise — it was a deliberate architectural decision driven by the nature of the data each product handles.

The Data Sovereignty Question

Organizational intelligence reasons over internal data: sprint boards, meeting notes, incident logs, decision records. This data is sensitive and often subject to compliance requirements. Sending it to a cloud API — even with encryption — creates a trust boundary that many enterprises aren't comfortable with. An SDK that runs inside their infrastructure eliminates this concern entirely.

When Cloud Makes Sense

Marketing intelligence analyzes public websites — data that's already accessible to anyone with a browser. There's no sovereignty concern, and the computational requirements (14 parallel agents, LLM calls, web crawling) are better served by elastic cloud infrastructure. SaaS also enables features like scheduled monitoring, team collaboration, and white-label reporting that would be complex to self-host.

Shared Principles, Different Models

Despite the different deployment models, both products share core principles: structured reasoning over raw retrieval, deterministic outputs, full auditability, and domain-specific AI rather than general-purpose chatbots. The delivery mechanism differs, but the engineering philosophy is the same.

The Multi-Product Advantage

Operating both models gives us a unique perspective. Lessons from building permission-aware reasoning for enterprises improve our SaaS agent design. Scale challenges from running 14 parallel agents in the cloud inform our SDK performance optimization. Each product makes the other better.