AI Frameworks · MemPalace

mempalace

MemPalace is a local-first AI memory system that stores conversation history verbatim and retrieves it using semantic search, achieving 96.6% recall without API calls. It organizes memories hierarchically (wings for entities, rooms for topics, drawers for content) and supports pluggable backends including ChromaDB, SQLite, Qdrant, and PostgreSQL+pgvector.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/MemPalace/mempalace

57.1k

GitHub stars

7.4k

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	MemPalace/mempalace
Owner	MemPalace
Primary language	Python
License	MIT — OSI-approved
Stars	57.1k
Forks	7.4k
Open issues	596
Latest release	v3.5.0 (2026-06-23)
Last updated	2026-07-07
Source	https://github.com/MemPalace/mempalace

What mempalace is

Python-based semantic search engine with structured indexing, pluggable vector storage backends, and MCP server integration. Stores raw text locally by default, offers optional hybrid retrieval with keyword and temporal boosting, and includes a SQLite-backed temporal knowledge graph with entity relationships and validity windows.

Quickstart

Get the mempalace source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/MemPalace/mempalace.git2cd mempalace3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Claude Code and LLM IDE Integration

Retrieve conversation history and project context within Claude Code sessions without 30-day expiration; wire via MCP stdio server for seamless agent memory.

Local-First Knowledge Retrieval for AI Applications

Build RAG systems and chatbots that store sensitive conversation data on-device; avoid cloud APIs and vendor lock-in while maintaining 96%+ semantic retrieval accuracy.

Session Context Mining from Development Workflows

Index project files, Claude conversations, and development logs; surface relevant prior decisions and rationales during new work without manual summarization.

Implementation considerations

Install in isolated environment (uv tool install, pipx, or venv) to avoid PEP 668 conflicts; ChromaDB and its deps (numpy, grpcio) can clash with system Python.
Mine content upfront (mempalace mine ~/path) to populate the palace; retrieval quality depends on indexed content volume and semantic relevance of stored text.
Choose a backend before heavy indexing: ChromaDB (default, in-process), sqlite_exact (local, exact-vector validation), qdrant (REST), or pgvector (Postgres). Migration between backends is not documented.
For Claude Code integration, configure MCP server and set auto-save hooks; README warns that sessions expire in 30 days without retention setup.
Embeddings model is cached locally; first run downloads it. Disk and memory footprint depend on backend and corpus size; unknown baseline requirements.

When to avoid it — and what to weigh

Real-Time, High-Throughput Vector Search — If you need sub-millisecond latency at millions of queries/day, the default ChromaDB setup is not optimized for that scale; external backends (Qdrant, pgvector) may help but require operational overhead.
Multi-Tenant SaaS without Self-Hosting — MemPalace is designed for single-user or small-team local deployment. Running as a managed service would require significant architectural changes and is not the project's current scope.
Mature, Stable Production Deployments Requiring Long-Term Support — Project created April 2026; version v3.5.0 is recent. Adopt only if you can tolerate API changes, breaking updates, or limited vendor stability guarantees.
No Dependency on External Embeddings — MemPalace relies on embeddings (ChromaDB bundles a model). If you cannot run or cache a local embedding model, you'll depend on third-party services or external APIs.

License & commercial use

MIT License. Permissive, OSI-approved; allows commercial use, modification, and distribution with attribution and liability disclaimer.

MIT License permits commercial deployment. However, project maturity (created April 2026, v3.5.0 current) is recent; no SLA, warranty, or long-term support guarantees stated. Use in production at own risk; recommend contractual review if embedding in customer-facing products.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Low
DEV.co fit	Good
Assessment confidence	High

Security considerations

Local-first by design; content does not leave device unless opted into external backends (Qdrant, pgvector), in which case verbatim text and metadata are sent to the configured service. README warns of impostor domains and recommends verifying only official PyPI and GitHub sources. No CVE history, penetration test results, or security audit mentioned. CAUTION notice about Claude Code 30-day session expiry suggests awareness of edge cases but no formal threat model published.

Alternatives to consider

Mem0 / Zep / Hindsight

Mature, cloud-first memory systems with managed hosting and multi-tenant support. Trade local control and latency for operational simplicity; README explicitly declines head-to-head benchmarking.

Chroma / Weaviate / Milvus

General-purpose vector databases without memory-system semantics. Require custom application code to structure memories (wings/rooms/drawers), manage indexing pipelines, and wire MCP. Better for high-scale, multi-tenant deployments.

Supermemory / Mastra

Lightweight alternatives with different retrieval metrics and architectural trade-offs. Less transparent benchmarking; MemPalace publishes reproducible results. Evaluate based on your domain (conversation vs. document memory).

Software development agency

Build on mempalace with DEV.co software developers

Start with MemPalace: install via uv, mine your first project, and integrate into Claude Code in minutes. Evaluate the pluggable backends for your scale.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

mem0 memsearch memvid

Related on DEV.co

Explore the category and the services that help you build with it.

AI Frameworks AI Application Development API Development Web App Development Talk to DEV.co

mempalace FAQ

Do I need an API key or internet connection to use MemPalace?

No for the default setup with ChromaDB and local embeddings. Semantic search (96.6% R@5) works entirely offline. External backends (Qdrant, pgvector) and optional LLM reranking require connectivity to those services, but are opt-in.

How do I migrate between storage backends?

Not clearly stated. Switching from ChromaDB to Qdrant or pgvector requires re-indexing (mempalace mine). No export/import tools or migration guide documented.

What are the hardware/disk requirements?

Unknown. Embeddings model is cached locally (size not specified). Index size depends on content volume and backend. No benchmarks for memory/CPU/disk footprint provided.

Is there a GraphQL or REST API instead of MCP?

Not mentioned. Primary interfaces are CLI (mempalace mine/search/wake-up) and MCP server (35 tools). Custom integration via Python API is possible (import mempalace) but undocumented in README.

Software developers & web developers for hire

Need help beyond evaluating mempalace? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and ai frameworks integrations — and maintain them long-term.

Ready to Build Smarter AI Memory?

Start with MemPalace: install via uv, mine your first project, and integrate into Claude Code in minutes. Evaluate the pluggable backends for your scale.

Talk to DEV.co Schedule a Consultation