“I want ChatGPT with infinite local memory” - Here’s how

MemAgent, a largely unknown project, enables AI agents like ChatGPT and Claude to have infinite, 100% local and private memory across multiple applications.

• MemAgent provides local, private memory for AI agents like ChatGPT and Claude.

• The setup involves cloning the GitHub repository, installing LM Studio, and configuring the agent.

• MemAgent can be used with various software through memory connectors for ChatGPT, GitHub, and Google Docs.

MemAgent provides infinite local memory for AI agents and connects memory between various applications, ensuring data privacy. It utilizes a 4 billion parameter LLM fine-tuned for memory management and a server (MCP) to expose its capabilities to other applications. The setup involves cloning the GitHub repository, installing LM Studio, and configuring the agent, which can be achieved even without extensive coding knowledge.

GitHub Repository

• 00:02:27 To begin, the GitHub repository must be cloned, which involves copying the provided link and using a code editor like Cursor or VS Code. Navigating to the correct directory without spaces in the folder names is crucial to prevent installation issues. The project uses a make file to simplify the installation process.

LM Studio Installation

• 00:03:40 The next step involves using the command 'make install' to install LM Studio, which is an application that allows the user to run AI models locally. Although some users might already have LM Studio with the UI version, this command installs the command-line interface, which allows projects and applications to be built without the graphical user interface.

Running MemAgent

• 00:05:26 To run the MemAgent, the command 'make run-agent' is used, which includes installing the 4 billion parameter model locally. There are three different models available depending on the computer's processing power: 4-bit quantized (fastest), 8-bit (balanced), and 16-bit (most powerful). The setup will prompt you to install the model from Hugging Face if it is not already installed.

Configuring Local Memory

• 00:07:32 A folder should be chosen to store local memory, where MemAgent will store all the markdown files, which can be manually edited. After creating a new terminal and navigating to the MemAgent MCP folder, the command 'make setup' is entered, which runs a Python script to select a folder to store the MD files.

Connecting to Claude Desktop

• 00:09:57 To connect MemAgent to Claude desktop, the 'make generate-mcp-json' command is used to create an mcp.json file, which is then integrated into the Claude desktop config.json file. A modified setup using UV is recommended for a cleaner execution, and after adjusting the configuration, Claude desktop is restarted, and the MCP server is initiated using 'uv run python mcp_server/server.py'.