Local-first by default
Agent context should be inspectable files and small databases before it becomes a remote service.
Memory infrastructure for AI agents
Give agents a place to remember from.
Substratia is a local-first memory layer for agent work: compact continuity records, portable context, and tools that keep future sessions from rebuilding the same understanding again and again.
What this is
A small substrate for durable agent context.
The public site is intentionally static. The product work lives in repo-owned packages, command-line tools, and plain files that can be audited without a web account or hosted dashboard.
Fast recovery
Continuity is useful when it makes the next session cheaper: fewer summaries, fewer forgotten decisions, fewer restarts.
Portable memory
Snapshots, prompts, and records should be easy to move between tools instead of being trapped in one chat surface.
System surface
Built around files, CLIs, and explicit handoffs.
Substratia favors source-controlled docs, local state, and compact savepoints over hidden services. That keeps the operational model legible for humans and cheap for deployment.
Thread savepoints for project continuity
A local memory SDK and CLI
Durable snapshots for handoffs and compaction
Security checks for API-key handling
No surprise billing posture
The website exports to static files.
No middleware, no serverless API routes, no hosted auth, no runtime image optimizer, and no analytics beacons. Vercel serves the built `out/` directory as static assets.