--- layout: home hero: name: MemPalace text: Give your AI a memory. tagline: "Local-first AI memory. Verbatim storage, pluggable backend, 96.6% R@5 raw on LongMemEval — zero API calls." image: src: /mempalace_logo.png alt: MemPalace actions: - theme: brand text: Get Started link: /guide/getting-started - theme: alt text: Architecture → link: /concepts/the-palace - theme: alt text: GitHub ↗ link: https://github.com/MemPalace/mempalace features: - icon: src: /icons/file-text.svg alt: Verbatim Storage title: Verbatim Storage details: Store source text directly instead of extracting facts up front. The raw benchmark result comes from retrieving verbatim content. - icon: src: /icons/building-2.svg alt: Palace Structure title: Palace Structure details: Wings and rooms give retrieval useful structure. In the project benchmarks, narrowing search scope outperformed flat search. - icon: src: /icons/search.svg alt: Semantic Search title: Semantic Search details: Vector search over verbatim content lets the model retrieve past discussions by topic, project, or room. Backend is pluggable. - icon: src: /icons/git-merge.svg alt: Knowledge Graph title: Knowledge Graph details: Temporal entity-relationship triples in SQLite. Facts can be added, queried, and invalidated over time. - icon: src: /icons/wrench.svg alt: 19 MCP Tools title: 19 MCP Tools details: MCP tools expose search, filing, knowledge graph, graph navigation, and diary operations to compatible clients. - icon: src: /icons/shield-check.svg alt: Zero Cloud title: Zero Cloud details: Core storage and retrieval run locally. Optional reranking features can add an API dependency but are not required for the benchmark path. ---
## Verbatim Retrieval First MemPalace stores source text and retrieves it with semantic search. The benchmarked raw mode does not require an LLM at any stage — no extraction, no rerank, no summarisation. **LongMemEval retrieval recall (500 questions):** | Mode | R@5 | LLM required | |---|---|---| | Raw (semantic search over verbatim text) | **96.6%** | None | | Hybrid v4, held-out 450q | **98.4%** | None | The raw 96.6% reproduces on any machine with the committed dataset: result JSONLs, the `seed=42` train/held-out split, and the `--mode raw` / `--held-out` runners are all in the `benchmarks/` directory of the repo. We deliberately do not publish a side-by-side comparison against other memory systems on this page. Retrieval recall (R@5) and end-to-end QA accuracy are different metrics and are not comparable; where MemPalace can be fairly compared on the same metric, we link to the other project's published source.
Full benchmark methodology →