Hojo Memory Engine Open-Sources Agent Memory
Hojo Memory Engine is an Apache-2.0, schema-first memory layer for AI agents. Its continuous meeting memory release tracks decisions, revisions, risks, and provenance across meetings, with policy-aware retrieval and human-gated self-improvement.
Hojo’s strongest idea is treating agent memory as governed application state—not a pile of retrieved text.
- –Atomic claims preserve provenance and revision history as project context evolves
- –Policy checks run before retrieval, improving control over sensitive or role-specific memories
- –Python and Java SDKs, dashboards, and deployment components target production agent teams
- –Human review gates memory corrections and rule changes, limiting self-improvement drift
- –The tradeoff is operational complexity: MySQL, MongoDB, Qdrant, Redis, Kafka, and Temporal make this heavier than lightweight memory middleware
DISCOVERED
1h ago
2026-10-05
PUBLISHED
1h ago
2026-10-05
RELEVANCE
AUTHOR
hojoHQ
