mempalace
SHA-256Open-source local-first AI memory system with verbatim storage and semantic search — zero API calls, no keys, top benchmark scores.
Smart Download
Download Download Version
v3.11.0 · 2.6 MB
Permanent, free, local-first long-term memory for your AI assistants.
Core Features
- Verbatim storage of conversation history — no summarization, extraction, or paraphrasing
- Structured palace-style index (wings/rooms/drawers) for scoped instead of flat search
- Local-first with zero API calls: data never leaves your machine unless you opt in
- Pluggable backends: ChromaDB by default, plus SQLite, Rust-native, Milvus, Qdrant, pgvector
- Built-in MCP server for Claude Code, Gemini CLI, and other coding agents
What It Can't Do
- •Beware impostor sites: the only official sources are the GitHub repo, the PyPI package, and mempalaceofficial.com — other domains may distribute malware. Claude Code sessions expire after 30 days unless you wire auto-save hooks. Native Android/Termux installs are unsupported (no Android wheels for ChromaDB/ONNX); use a Debian PRoot container instead. The first run needs network access to download an ~80–300MB embedding model. On Linux, bind-mounted directories must be readable by uid 1000 or you will hit PermissionError; the GPU image is x86_64-only and fails to build on Apple Silicon. Finally, it is a retrieval tool — it never summarizes or extracts knowledge, so look elsewhere if you need automatic digests.
Use Cases
- Give Claude Code and other AI coding assistants persistent cross-session memory
- Ask natural-language questions about past decisions, e.g. "why did we switch to GraphQL"
- Run a shared-brain hub so a whole team queries one collective memory
- Manage AI memory in fully offline, privacy-sensitive environments
MemPalace is the best-benchmarked open-source AI memory system, scoring 96.6% R@5 on LongMemEval. It stores conversation history as verbatim text — no summarizing, extracting, or paraphrasing — and retrieves it with semantic search. The index is structured like a palace: people and projects become wings, topics become rooms, and original content lives in drawers, so searches can be scoped instead of scanning a flat corpus. The retrieval layer is pluggable, with ChromaDB as the default and opt-in backends for SQLite, Rust-native vectors, Milvus, Qdrant, and pgvector. Compared to cloud-oriented memory tools like Mem0 or Zep, MemPalace is local-first and makes zero API calls: nothing leaves your machine unless you opt in, there are no usage fees, and no API keys are required. It ships a CLI, an MCP server for Claude Code and other agents, multi-arch Docker images, and team hub modes for shared-brain setups. MIT licensed.
Tags
Getting Started
Download installer
Click the button above to download the installer for your system
Install the software
Double-click the downloaded installer and follow the prompts
Install uv, then run: uv tool install mempalace (pipx install mempalace works too)
Run mempalace init ~/projects/myapp to initialize a palace in your project
Run mempalace mine ~/projects/myapp to index project files or conversation transcripts
- Install uv, then run: uv tool install mempalace (pipx install mempalace works too)
- Run mempalace init ~/projects/myapp to initialize a palace in your project
- Run mempalace mine ~/projects/myapp to index project files or conversation transcripts
SHA-256 checksum verified
Checksum extracted from GitHub official Release page
SHA256 Checksum
ecf5f8dec80cf2f07a2d33dbebb3d323f3872cc7b57a6cd1fa446dfd3978f415This checksum is extracted from the GitHub Release page. Verify file integrity after download.
All SHA-256 checksums on this platform are extracted from the project's official GitHub Release page, without any modification. You can independently verify them on the GitHub Releases page.
Open Source Transparency
View GitHub SourceUninstall Info
Run uv tool uninstall mempalace or pipx uninstall mempalace to remove the CLI, then delete the palace data directory and config.json created during init. Docker users should also remove the image (docker rmi) and the volume (docker volume rm mempalace-data). If you wired it into an MCP client, remove the mempalace entry from the client configuration as well.
No Extra Dependencies
Ready to use after download. No additional runtime required.
Having issues? Check the FAQ below
3 FAQs
Similar Projects
ollama
Ollama lets you download, run, and manage large language models locally. One command, multiple platforms, endless possibilities.
llama.cpp
High-performance LLM inference engine in C/C++ with minimal dependencies, supporting quantized models (1.5–8 bit) and diverse hardware (Apple Silicon, CUDA, Vulkan, etc.).
gemini-cli
Google's open-source terminal AI agent for coding, debugging, and workflow automation powered by Gemini models.