Built on a simple principle.
Only use an LLM when it is absolutely necessary. Most of the work thrown at cloud services by conventional AI tools can be done far more efficiently - and more cheaply - through local, deterministic processes. We build the infrastructure that enforces that principle on your behalf.
Sedasoft
Hydrate is a product of Sedasoft, a software company created by two industry veterans who have approached the challenge of building more efficient and trustworthy AI tools from the context of a combined 50+ years of experience in information and data architecture.
In the rush to embrace AI, the principles of sound data architecture have been discarded in the mistaken belief that you can simply throw unstructured information at an LLM and get value out the other side. We built Sedasoft to prove the alternative: get the structure of your information right first, and the AI becomes a reliable multiplier rather than an expensive gamble.
What Hydrate is built on
Hydrate is a commercial product built on top of siteengine_ai, the memory infrastructure that powers it. siteengine_ai is the engine; Hydrate is the developer product around that engine. We ship both because they solve different problems. siteengine_ai is the lower-level memory substrate that any application can use; Hydrate is the integrated experience for Claude Code and MCP-capable agents.
If you are asking whether Hydrate is "just a wrapper around siteengine_ai" — the honest answer is yes, and that is load-bearing for the value claim. The engine is the differentiated part. Hydrate adds the hook integration, the dashboard, the team-canon manifest, the licence layer, the install path, and the on-call story; the memory engine underneath has been in development since well before the current memory-tool wave.
LLMs last, not first
We never call an LLM where a faster, cheaper, deterministic method will do. Fact deduplication, session summarisation, and context ranking all happen locally - without API calls. LLM extraction is the exception, not the rule.
Your data stays yours
Hydrate runs local-first by design. Your coding sessions, your architectural decisions, your project context - none of it leaves your machine unless you explicitly choose Enterprise sync. No telemetry. No training. No cloud custody.
Architecture before intelligence
Getting the structure of your information right matters more than which model you query it with. Hydrate's architectural canon feature forces that rigour onto AI coding sessions - pinned rules that no prompt can override.
Get in touch
Press, partnership, and enterprise enquiries: [email protected]
For technical support and bug reports, open an issue on GitHub.