mempalace

What it is and what it does

MemPalace is a local-first AI memory system that stores conversation history and project files verbatim and retrieves them using semantic search. It does not summarize or paraphrase—it preserves the original text and indexes it into a structured hierarchy (wings for people/projects, rooms for topics, drawers for content), allowing scoped searches rather than flat corpus queries. The retrieval layer is pluggable; ChromaDB is the default, but you can swap in Qdrant, Milvus, pgvector, or SQLite depending on your deployment model.

It runs entirely on your machine with no API calls or external services required. The package ships both as a CLI tool (best installed via uv tool install or pipx to isolate dependencies) and as a Python library. It integrates with Claude Code and MCP-compatible tools. The fact sheet shows 7 runtime dependencies (chromadb, huggingface-hub, numpy, python-dateutil, pyyaml, tokenizers, tomli), all well-established packages.

Use it for:

• Mine Claude Code session transcripts and project files into a searchable palace, then retrieve context for new sessions without manual summarization.
• Build a searchable archive of past conversations and decisions for a team or long-running project, scoped by wing (project) and room (topic).
• Integrate MemPalace as an MCP server into Claude Code or other compatible tools to give the AI access to your project history and conversation context.
• Deploy a local or server-backed vector search layer (via Qdrant, Milvus, or pgvector) for retrieval-augmented generation without cloud dependencies.
• Use the CLI to mine codebases and documentation into a palace, then query it from the command line or via the Python library.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

MemPalace stores conversation history and project files as verbatim text and retrieves them via semantic search, with a structured index (wings, rooms, drawers) and pluggable backends including ChromaDB, Qdrant, Milvus, and pgvector.

Yes. MemPalace is actively maintained (release 2026-08-14), has no known vulnerabilities, uses MIT license, and solves a real problem—giving AI tools access to your conversation and project history without API calls. Low install friction via uv/pipx, and the pluggable backend design means you can start with ChromaDB and migrate later. Install it if you work with Claude Code, need local RAG, or want to preserve and search your conversation history.

Install

mempalace on PyPI

pip install mempalace

uv add mempalace

poetry add mempalace

Installing mempalace

Before you install

Low friction: pure Python wheel, ships as a CLI tool best installed via uv tool install or pipx to avoid dependency conflicts. Active maintenance with release on 2026-08-14 and 58372 GitHub stars.

License in practice

MIT license permits commercial and private use with minimal restrictions—suitable for both personal and production deployments.

Quickstart

# Install via uv (recommended for CLI isolation)
uv tool install mempalace
mempalace init ~/projects/myapp
mempalace mine ~/projects/myapp
mempalace search "why did we switch to GraphQL"

# Or in a virtualenv for library use:
pip install mempalace
import mempalace

First embedding model download (~80 MB for default minilm, ~300 MB for embeddinggemma) requires network access and is cached thereafter. Requires Python >=3.9.

Verify before relying

• Whether the 96.6% R@5 benchmark on LongMemEval is reproducible in typical user workflows or specific to the test dataset.
• Performance characteristics and memory footprint when indexing large codebases or multi-year conversation histories.
• Stability guarantees for the pluggable backend interface across minor version updates.

Package facts

Evidence: mempalace-3.7.1-py3-none-any.whl

Keywords: ai, chatgpt, chromadb, claude, embeddings, llm, mcp, memory, rag, vector-database

Tags

More Utilities packages