Anamnesis documentation

Anamnesis is a persistent, provenance-aware memory layer for AI tooling: it captures working sessions automatically, consolidates them into trustworthy long-term memory, and serves the current answer back to any connected model. Everything here is written to the same standard as our benchmarks: say what's true, link the evidence, state what we don't claim.

Start here

• Quickstart — from zero to persistent memory in about ten minutes: the Claude Code plugin for the full capture + recall loop, or any MCP client via your personal connector.
• Product Datasheet — what it does, what it replaces, integrations, and deployment tiers, in buyer language.
• Executive Summary — the one-page version: problem, product, measured results, moat. Also as a PDF.

Trust & architecture

• Trust Center — security posture, data ownership, sub-processors, and what we deliberately do not claim.
• Data Sovereignty — the tier ladder from hosted to fully on-premises, and the one-way mirror.
• Security Architecture — the mechanisms: read-only guest access, IP containment, provenance trust tiers.

The evidence

Every quantitative claim in these pages traces to the benchmark suite — pre-registered, reproducible, blind-graded, with honest negatives reported alongside the wins.