ElBruno.MempalaceNet
mcp
Warn
Health Warn
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 5 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
Purpose
This tool provides a local-first semantic memory system for AI applications and agents. It ingests and stores content locally using SQLite and ONNX embeddings, and exposes this knowledge via an MCP server so that compatible AI clients can perform semantic searches.
Security Assessment
Overall Risk: Low. The project operates with a "local-first" design, meaning it does not make external cloud calls or require API keys by default. The automated code scan of 12 files found no dangerous patterns, no hardcoded secrets, and no dangerous permission requests. Because it functions as a data ingestion tool, it will naturally read and access local files and folders based on the directories you point it to. While it does not natively execute arbitrary shell commands or make external network requests, users should be aware that it relies on an SQLite backend to store a verbatim copy of the mined data locally.
Quality Assessment
The project is in active development and shows strong signs of good engineering health. It is openly licensed under the highly permissive MIT license. The creator maintains a continuous integration pipeline, and the README reports that all 152 unit tests are currently passing. However, community visibility and overall public adoption are currently very low, as indicated by its 5 GitHub stars. Consequently, the tool has not undergone widespread peer review from the broader developer community.
Verdict
Safe to use, though you should keep in mind that it is a young project with limited public community vetting.
This tool provides a local-first semantic memory system for AI applications and agents. It ingests and stores content locally using SQLite and ONNX embeddings, and exposes this knowledge via an MCP server so that compatible AI clients can perform semantic searches.
Security Assessment
Overall Risk: Low. The project operates with a "local-first" design, meaning it does not make external cloud calls or require API keys by default. The automated code scan of 12 files found no dangerous patterns, no hardcoded secrets, and no dangerous permission requests. Because it functions as a data ingestion tool, it will naturally read and access local files and folders based on the directories you point it to. While it does not natively execute arbitrary shell commands or make external network requests, users should be aware that it relies on an SQLite backend to store a verbatim copy of the mined data locally.
Quality Assessment
The project is in active development and shows strong signs of good engineering health. It is openly licensed under the highly permissive MIT license. The creator maintains a continuous integration pipeline, and the README reports that all 152 unit tests are currently passing. However, community visibility and overall public adoption are currently very low, as indicated by its 5 GitHub stars. Consequently, the tool has not undergone widespread peer review from the broader developer community.
Verdict
Safe to use, though you should keep in mind that it is a young project with limited public community vetting.
MemPalace.NET — local-first AI memory in .NET (port of MemPalace)
README.md
MemPalace.NET
A .NET port of MemPalace — local-first AI memory that stores everything verbatim, searches semantically, and organizes knowledge through a wings / rooms / drawers hierarchy. No cloud calls by default, powered by ONNX embeddings.
🎯 Status: v0.6.0 — Production-ready with Copilot Skill infrastructure and full semantic memory capabilities.
Why MemPalace.NET?
- Local-first by default — ONNX embeddings via ElBruno.LocalEmbeddings (no API keys, no cloud calls)
- Microsoft.Extensions.AI — swap embedders and LLMs with zero lock-in
- Microsoft Agent Framework — each agent gets its own memory diary
- MCP server — expose your palace as Model Context Protocol tools (Claude Desktop, VS Code, etc.)
- Temporal knowledge graph — track entity relationships with validity windows
- SQLite backend — managed BLOB storage, cosine similarity, clear upgrade path to vector stores
Examples & Getting Started
Ready to dive in? Check out our runnable examples:
- 🔰 Simple Memory Agent — Core memory operations with semantic search
- 🕸️ Semantic Knowledge Graph — Temporal entity relationships
See examples/README.md for detailed walkthroughs and learning paths.
Quick Start
# Install the CLI tool
dotnet tool install -g mempalacenet --version 0.6.0
# Initialize a new palace
mempalacenet init ~/my-palace
# Mine project files
mempalacenet mine ~/my-code --wing work --mode files
# Mine conversation transcripts
mempalacenet mine ~/my-convos --wing personal --mode convos
# Semantic search
mempalacenet search "how do I handle auth errors?"
# Hybrid search with reranking
mempalacenet search "latest React patterns" --hybrid --rerank
# Start MCP server (for Claude Desktop, VS Code, etc.)
mempalacenet mcp --palace ~/my-palace
# Run an agent
mempalacenet agents run scribe --wing research --mode local
Architecture
MemPalace.NET is a modular .NET solution with clear separation of concerns:
| Project | Purpose |
|---|---|
| MemPalace.Core | Domain types, storage interfaces, PalaceRef value object |
| MemPalace.Backends.Sqlite | Default SQLite backend with BLOB vectors + cosine similarity |
| MemPalace.Ai | M.E.AI integration with ONNX (default), Ollama, OpenAI support |
| MemPalace.Mining | Content ingestion: filesystem miner + conversation transcript miner |
| MemPalace.Search | Semantic, keyword, and hybrid search with optional LLM reranking |
| MemPalace.KnowledgeGraph | Temporal entity-relationship graph with validity windows |
| MemPalace.Mcp | Model Context Protocol server (7 tools in v0.1) |
| MemPalace.Agents | Microsoft Agent Framework integration + per-agent diaries |
| MemPalace.Cli | Spectre.Console CLI (mempalacenet command) |
| MemPalace.Benchmarks | LongMemEval / LoCoMo / ConvoMem benchmarks + R@5 testing |
Documentation
Full documentation lives in docs/:
- Architecture — solution layout, component contracts, dependency graph
- Concepts — wings, rooms, drawers, verbatim storage, embedder identity
- Backends — writing custom backends, conformance tests
- AI Integration — embedder selection, reranking, M.E.AI seams
- Mining — ingestion pipeline, custom miners, .gitignore respect
- Search — semantic vs hybrid strategies, RRF fusion, temporal boosting
- Knowledge Graph — temporal triples, pattern queries, invalidation
- MCP Server — tool reference, VS Code / Claude Desktop setup
- Agents — Agent Framework integration, diary management, agent discovery
- CLI — command reference, configuration, examples
- Benchmarks — reproducibility, dataset sources, R@5 parity
- GitHub Copilot Skill — integration guide, pattern library, code generation hints
Building Custom Integrations
Integrating MemPalace.NET into your .NET projects? Start with our developer guides:
- C# Library Developer Guide — Build applications on MemPalace.NET
- Embedder Pluggability Guide — Swap or implement custom embedders
- Skill Integration Deep Dive — Extend the platform with reusable skills
Development
# Clone
git clone https://github.com/elbruno/mempalacenet
cd mempalacenet
# Build
dotnet build src/
# Test (129 tests, all green)
dotnet test src/
# Pack NuGet packages
dotnet pack src/ -c Release
Roadmap
v0.1.0 (current) ships core memory operations, search, MCP server, and agents.
Post-v0.1:
- Upgrade to sqlite-vec or Qdrant for >100K vectors
- BM25 keyword search (currently token overlap)
- LongMemEval R@5 parity validation (target ≥ 91%)
- Conversation context summaries (
mempalace wake-up)
Credits
- Original project: MemPalace (Python)
- Default embedder: ElBruno.LocalEmbeddings (ONNX)
License
MIT — same spirit as the original MemPalace.
👋 About the Author
Made with ❤️ by Bruno Capuano (ElBruno)
- 📝 Blog: elbruno.com
- 📺 YouTube: youtube.com/elbruno
- 🔗 LinkedIn: linkedin.com/in/elbruno
- 𝕏 Twitter: twitter.com/elbruno
- 🎙️ Podcast: notienenombre.com
Community
We welcome contributions from the community! Here's how to get involved:
- Contributing Guidelines — how to submit pull requests, report issues, and contribute code
- Code of Conduct — our commitment to fostering an open and welcoming environment
- Security Policy — how to report security vulnerabilities responsibly
- Issues — report bugs or request features
- Discussions — ask questions, share ideas, and connect with the community
Got questions? Open a discussion or reach out to @elbruno.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found