Open Source — Free to Use

MemPalace

Local-first AI memory. Verbatim storage, pluggable backend, and 96.6% R@5 raw recall on LongMemEval — with zero API calls. Open-source and free to use.

mempalace — benchmark
# Install MemPalace
$ pip install mempalace
 
# Run LongMemEval benchmark (500q, no LLM)
$ python benchmarks/longmemeval_bench.py
 
Loading 500 questions...
Semantic search — no heuristics, no rerank...
 
✓ R@5 (raw): 96.6%
✓ R@5 (hybrid v4, held-out): 98.4%
96.6%
LongMemEval R@5 (raw)
29
MCP Tools
0
API Keys Required
MIT
Open Source License

What Makes MemPalace Different

A memory architecture designed from first principles — not an afterthought bolted onto existing systems.

🧠

Verbatim Storage

Conversations are stored word-for-word — not summarized, extracted, or paraphrased. 96.6% R@5 raw on LongMemEval, 98.4% with the hybrid pipeline.

🏛

Structured Palace

People and projects become wings, topics become rooms, content lives in drawers — so searches can be scoped, not run against a flat corpus.

Temporal Knowledge Graph

A local SQLite-backed entity-relationship graph with validity windows. Add, query, invalidate, and timeline facts as they evolve.

🔒

Local-First

Nothing leaves your machine unless you opt in. No API keys, no cloud, no LLM required for retrieval. MIT-licensed and fully inspectable.

🔌

Pluggable Backend

ChromaDB by default, but the retrieval interface is clean — swap in an alternative backend without touching the rest of the system.

🔗

29 MCP Tools

Palace reads/writes, knowledge-graph operations, cross-wing navigation, drawer management, and agent diaries — ready for Claude Code, Gemini CLI, and any MCP-compatible tool.

Architecture

MemPalace uses a layered memory architecture inspired by human cognition — separating working memory, episodic memory, and long-term semantic storage into distinct but interconnected systems.

  1. 1
    Ingestion Layer Processes incoming information through semantic parsing, entity extraction, and relationship mapping.
  2. 2
    Memory Palace The core storage engine using a novel spatial-semantic hybrid index for instant retrieval with contextual understanding.
  3. 3
    Recall Engine Multi-modal retrieval system that combines vector similarity, temporal proximity, and relational graph traversal.
  4. 4
    Context Synthesizer Assembles retrieved memories into coherent context windows, respecting recency, relevance, and token budgets.
INPUT STREAM
Conversations, documents, structured data
INGESTION LAYER
Semantic parsing • Entity extraction • Relationship mapping
MEMORY PALACE CORE
Spatial-semantic hybrid index • Temporal versioning • Graph store
VECTOR SEARCH
Embedding similarity
GRAPH WALK
Relational traversal
TEMPORAL
Time-aware recall
CONTEXT SYNTHESIZER
Assembles coherent context • Token budget management • Relevance ranking
OUTPUT
Rich context for any LLM • 96.6% R@5 raw on LongMemEval

Benchmark Results

Every number below is reproducible from the public repo. Full per-question result files are committed under benchmarks/results_*.

Benchmark Metric Score Notes
LongMemEval — raw R@5 96.6% 500 questions • no LLM, no heuristics
LongMemEval — hybrid v4 R@5 98.4% 450q held-out (tuned on 50 dev)
LongMemEval — hybrid v4 + rerank R@5 ≥99% Any capable LLM as reader
ConvoMem Avg recall 92.9% 250 items, 50 per category
LoCoMo — hybrid v5 R@10 88.9% 1,986 questions, no rerank
MemBench (ACL 2025) R@5 80.3% 8,500 items, all categories

MemPalace deliberately does not headline a “100%” number: the last 0.6% was reached by inspecting specific wrong answers, which the project flags as teaching to the test. 96.6% R@5 raw and 98.4% hybrid held-out are the honest, generalisable figures.

Built by Humans, for AI

MJ

Milla Jovovich

Architect

Actress, musician, model — and the architectural vision behind MemPalace. Designing how AI should remember.

BS

Ben Sigman

Engineer

Building the core engine, benchmarks, and infrastructure that makes perfect recall a reality.

Give Your AI
Perfect Memory

MemPalace is free, open-source, and ready to integrate. Whether you're building a chatbot, an agent, or something entirely new — give it the memory it deserves.