Robust Long‑Term Memory

Robust Long‑Term Memory is an MCP server that provides a persistent, human-like memory system for local AI companions running in clients like LM Studio. Built using a hybrid backend that pairs SQLite for structured metadata with ChromaDB for vector-based semantic retrieval, it allows models to retain facts, preferences, and events across sessions. Developers and users running local large language models use this server to eliminate the boundary of single-chat context windows. The server implements biological memory dynamics, applying idle-time lazy decay to unused memories while reinforcing memories that are frequently recalled. It supports automatic rotation of backups, JSON exports for cross-machine migrations, and continuous operation regardless of whether the user switches between different model checkpoints. The tool functions in the background, allowing the host assistant to query, record, update, or prune personal context without breaking conversational character.

Category: AI Memory & Context

Tags: context, memory, persistence, retrieval

Visit Robust Long‑Term Memory

How to install and configure Robust Long‑Term Memory

1. Clone the repository using git clone https://github.com/Rotoslider/long-term-memory-mcp.git and open the cloned folder. 2. Install required dependencies with pip install -r requirements.txt (optionally install pip install "huggingface_hub[hf_xet]"). 3. Register the server in your LM Studio mcp.json file: json { "mcpServers": { "long_term_memory": { "command": "python", "args": ["/path/to/long_term_memory_mcp/LongTermMemoryMCP.py"] } } } 4. Open LM Studio Server settings and load the long_term_memory tool.

What you can do with Robust Long‑Term Memory

• Storing personal user details, preferences, and conversation facts invisibly during ongoing chats. * Retrieving contextually relevant background information using vector semantic similarity searches. * Filtering and viewing stored memories by designated tags or specific classification types. * Reviewing recent discussions using temporal queries and date range filtering tools. * Creating manual snapshots and automated JSON backups across SQLite and ChromaDB files.

Key facts

• https://github.com/Rotoslider/long-term-memory-mcp
• AI Memory & Context
• context, memory, persistence, retrieval

Part of MCP Servers

What is Robust Long‑Term Memory?

Robust Long‑Term Memory is an MCP server that provides persistent storage and semantic retrieval for AI companions. It utilizes SQLite for relational metadata and ChromaDB for vector similarity searches, enabling language models to remember facts across different chat sessions and model changes.

Which MCP clients work with Robust Long‑Term Memory?

The server is documented specifically for LM Studio using its MCP configuration file. Because it adheres to the Model Context Protocol standard and runs via Python, it can also be adapted to other MCP-compatible hosts like Claude Desktop.

How does memory decay work in this server?

The server uses a human-like memory dynamic called lazy decay. Importance ratings decrease when a memory is accessed after long periods of inactivity, while memories that are retrieved frequently receive reinforcement to stay relevant.

How are backups handled?

Backups run automatically every 24 hours or after every 100 new entries, retaining the last 10 versions. Each backup contains copies of the SQLite database, ChromaDB data, and a portable JSON export.