MemPalace
Spatial-metaphor memory (wings / rooms / halls) on top of verbatim chunked storage. Independent analysis suggests the score is driven by verbatim storage + ChromaDB defaults rather than the palace structure.
At a glance
- Type
- Verbatim chunks + spatial metaphor
- Tier
- T2
- Section
- Research / specialised systems
- Created
- 2026-04
- Latest release
- v3.3.4 2026-05-01
- License
- MIT
- GitHub
- 51.3k★ +49.6k/mo Python
- Pricing
- searched not found
- Funding
- not applicable — not commercial
Taxonomy
- storage
- vector
- retrieval
- similarity
- persistence
- long-term
- update
- append-only
- unit
- chunk
- governance
- inspectable
- conflict
- append
When to use
Optimised for: research positioning (see memory_model)
Anti-fit: not applicable - research paper
Pros & cons
Pros
Three-layer corpus model (raw / governed / substrate) with explicit governance phases — most architecturally principled OSS memory framework.
Cons
OSS framework; less polished than commercial memory layers; smaller user base.
Claims & capabilities
Reports 96.6% Recall@5 on LongMemEval (later updated to 98.4% on held-out data after benchmark-tuning controversy). 47k★ in two weeks at launch.
Technical surface
- API surface
- searched not found
- Backend storage
- searched not found
- Deployment
- searched not found
- Embedding model
- searched not found
- Multi-tenancy
- searched not found
- MCP
- no first-party MCP adapter published as of 2026-05; community connectors may exist.
- A2A
- no Google A2A (Agent2Agent) integration documented as of 2026-05.
- OpenTelemetry
- no first-party OpenTelemetry exporter documented; standard logs/metrics typically available.
Similar systems
Other research / specialised systems in the catalog, ranked by inbound references.
- A-MEM T3
Treats memories as atomic linkable notes — explicit nod to Zettelkasten knowledge management. Dynamic linking; retroactive memory revision.
- BAI-LAB MemoryOS T3
Hierarchical "OS" with Storage / Updating / Retrieval / Generation modules. Short-term → mid-term via FIFO dialogue-chain; mid-term → long-term via segmented paging.
- Titans (Google) T4
Neural long-term memory module that learns to memorise at test time. Uses gradient-of-loss as "surprise" signal; adaptive weight-decay forgetting. Three variants: MAC (memory-as-context), MAG (memory-as-gate), MAL (memory-as-layer).
- EverMemOS T4
Self-organizing memory OS for structured long-horizon reasoning. Three-phase model: episodic, semantic, reconstructive.
- EVOLVE-MEM T3
Dynamic Memory Network + Hierarchical Memory Manager + Self-Improvement Engine. L0 raw embeddings, L1 contextual summaries, L2 high-level principles. NeurIPS 2025 (Scaling Environments for Agents workshop).
- LiCoMemory T4
Lightweight hierarchical graph (CogniGraph) with entities and relations as semantic indexing layers. Incremental graph construction, fast updates, low-latency inference. Nov 2025.
Related systems
References (1)
- Chroma builds on — Independent analysis suggests the score is driven by verbatim storage + ChromaDB defaults rather than the palace structure.