LORE
Lots Of Reconciled Engrams. Shared graph memory for Claude Code, Codex and DOXA, combining Hermes-pattern curated memory with Honcho-inspired dialectic reasoning.
Keep knowledge.
Keep its provenance.
LORE combines two memory patterns. Hermes-style curated memory keeps capped user and project facts, reviewable proposals and stable session snapshots. The Honcho-inspired belief layer derives conclusions, reconciles duplicates and contradictions, and supports dialectic reasoning over the evidence.
Beliefs retain their citations, confidence and typed relations, such as dependencies and contradictions. On-demand dialectic queries gather beliefs, curated facts and session evidence to produce cited answers. Curated memory writes go through review.
Hermes-pattern memory
Capped user and project facts, stable context snapshots, pending proposals and explicit approval gates for curated writes.
Dialectic reasoning
Derive evidence-linked beliefs, reconcile contradictions and ask for cited answers over the shared belief graph and local session history.
Shared graph memory
One local memory and belief store across Claude Code, Codex and DOXA, with typed graph relations, SQLite session search and optional signed sync. Host-specific facts stay local.
Use it with your agents.
The repository covers installation, platform support, integration and data boundaries.
/plugin marketplace add docwilde/lore
/plugin install lore
/lore:setup