Feature matrix
ValorBrain vs open-source memory
Honest matrix against mem0, MemPalace, supermemory, and Honcho. Product-ready features, not generic claims.
Disclosure: ValorBrain incorporates patterns from published papers and open-source projects (QMD, A-MEM, MAGMA, Honcho, MemPalace, supermemory, and others). This table compares usable product surfaces, not isolated components.
Updated August 28, 2026. Spot an error? [email protected]
| Feature | ValorBrain | mem0 | MemPalace | supermemory | Honcho |
|---|---|---|---|---|---|
| Storage | |||||
| Multi-tenant via Postgres RLS | yes | — | — | — | — |
| Self-host available | yes | partial | yes | partial | partial |
| On-prem option (zero third parties) | yes | — | partial | partial | enterprise |
| Vector backend | pgvector | configurable | sqlite-vec | configurable | managed |
| Retrieval | |||||
| Hybrid (BM25 + vector + RRF) | yes | — | yes | partial | partial |
| Cross-encoder reranking | yes | — | yes | — | partial |
| Intent-aware search (WHY/WHEN/ENTITY) | yes | — | yes | — | reasoning tier |
| Multi-graph traversal (semantic + temporal + causal) | yes | — | partial | — | partial |
| Multi-topic query decomposition | yes | — | — | — | — |
| Mechanics | |||||
| Self-evolving memory notes | yes | partial | yes | — | yes |
| Causal inference between observations | yes | — | partial | — | reasoning tier |
| Contradiction detection + supersedes | yes | — | partial | — | reasoning tier |
| Cross-session deductive synthesis | yes | — | — | — | reasoning tier |
| Background dreaming/consolidation | roadmap | — | partial | — | yes |
| Canonical sources + drift detection | yes | — | — | — | — |
| Cross-agent | |||||
| Multi-agent coherence (Claude/Cursor/Kiro) | yes | limited | limited | limited | peers |
| Automatic cross-session handoffs | yes | — | yes | partial | partial |
| Real-time multi-agent activity feed | yes | — | — | — | — |
| Integration | |||||
| Claude Code hooks (90% retrieval auto) | yes | — | yes | yes | via Skill |
| Public MCP HTTP transport | yes | — | stdio | stdio | — |
| OAuth 2.1 dynamic client registration | yes | — | — | — | — |
| Native OpenClaw plugin | yes | — | — | — | yes |
| Native Hermes Agent plugin | yes | — | — | — | yes |
| OAuth connectors (Notion, Drive, Slack…) | shipping soon | — | — | partial | — |
| Domain | |||||
| Multi-domain schemas (coding, marketing, legal…) | yes | — | — | — | — |
| Per-tenant schema customization | yes | — | — | — | — |
| Peer modeling (NPCs, multi-entity) | — | — | — | — | yes |
| Operations | |||||
| Unlimited documents and search on paid plans | yes | request quota | yes | yes | unlimited context() |
| Priced per person, not per API call | yes | — | self-host | — | — |
| Team product: seats, roles, invites | yes | — | — | — | — |
| Per-person permission on every document | yes | — | — | — | — |
| Audit logs + LGPD compliance | yes | partial | — | partial | partial |
| Data in Brazil + on-prem option | yes | — | your infra | partial | enterprise |
| Permission scopes per API token | yes | — | — | — | — |
| SaaS admin dashboard | yes | yes | — | yes | yes |
| Public benchmarks (LoCoMo, BEAM-100K) BEAM run 2026-08-23: end-to-end with GLM-5.3 Flash (Ox Alpha), judge GLM-5.2, effort=max, and about 44% more latency than default effort. This is a benchmark setting, not the chat default. Methodology → | 96.58% R@10 LoCoMo · 80.8% binary BEAM-100K | partial | — | — | yes |
available: yes/simmissing: —text = detail
Still deciding?
Minutes to set up. Free plan forever, no card. Compare on your stack.