Robust Long‑Term Memory
About
A persistent, human‑like memory system for AI companions
Details
- Author
- rotoslider
- Categories
- Other, Developer Tools
Jump to
Setup
Install Robust Long‑Term Memory in your MCP client (Claude Desktop, Cursor, Windsurf, and others).
Repository: https://github.com/rotoslider/long-term-memory-mcp
Follow the installation instructions in the repository README, then restart your MCP client.
Robust Long‑Term Memory MCP for LM Studio
A persistent, human‑like memory system for AI companions inLM Studio, powered by a hybrid of SQLite (structured storage) and ChromaDB (semantic search). It’s designed for decades‑long use, seamless recall across sessions, and automatic backups — making your AI companion feel like a continuous, living persona. Now with biological behavior: time‑based lazy decay and reinforcement by use.
- SQLitefor structured metadata and fast queries
- ChromaDBfor semantic similarity and natural recall
- JSON backupsfor portability
Cross‑chat continuity: memories persist beyond a single chat
Cross‑model continuity: swap models freely, the memory stays intact
Cross‑machine portability: move the database to another system and continue seamlessly
Automatic backups: daily backups and after every 100 memories, pruned to keep the last 10
Invisible memory integration: tools are hidden from the user; conversations feel natural
- Lazy Decay: importance decreases only when a memory is accessed after idle time
- Reinforcement: frequent recall strengthens memory importance
- Adaptive Semantic Threshold: balances precision/recall with a safe top‑1 fallback
- Clone the repo:
git clone https://github.com/Rotoslider/long-term-memory-mcp.git cd long-term-memory-mcp
chromadb sentence-transformers fastmcp (sqlite3 is built into Python; do not install separately)
3. (Optional) For faster HuggingFace model fetching:
pip install "huggingface_hub[hf_xet]"
Edit your LM Studio mcp.json to include the correct path:
{ "mcpServers": { "long_term_memory": { "command": "C:\\Python313\\python.exe", "args": [ "D:\\a.i. apps\\long_term_memory_mcp\\LongTermMemoryMCP.py" ], "env": {} } } }
- Open Server (MCP) Settings
- Load the MCP Tool: "long_term_memory"
- Cross‑Chats→ Start a new chat — memories are still there.
- Cross‑Models→ Switch models — the same memory remains available.
- Cross‑Machines→ Copy the database folder (memory_db/ and memory_backups/) and your system prompt, point to the path, and everything carries over.
💡 Think of it as your AI’s diary: chats are conversations, the database is the journal.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
