Memory-Plus

Memory-Plus is an MCP server that provides a lightweight, local Retrieval-Augmented Generation (RAG) memory store for AI agents across different coding sessions. It connects to the Gemini Embedding API using a Google API key to generate embeddings and stores persistent context locally on the developer's machine. Software engineers and developers working with coding tools like Cursor, Windsurf, Copilot, or Cline use this server to eliminate repetitive prompt setup. The server enables AI clients to record user notes, retrieve past entries via semantic keyword or topic searches, fetch recent entries, update items with full version history, and ingest documents via file import. Additionally, Memory-Plus provides graph clustering tools to visualize relationships between saved memory items, while using server resources to guide the client on when past interactions should or should not be recalled.

Category: AI Memory & Context

Tags: context, local, memory, persistence, rag

Visit Memory-Plus

How to install and configure Memory-Plus

1. Obtain an API key from Google AI Studio and set it as an environment variable named GOOGLE_API_KEY. 2. Ensure uv is installed on your system using pip install uv. 3. Add the server configuration to your client's settings file (such as settings.json in VS Code or Cursor, or cline_mcp_settings.json for Cline): json { "mcpServers": { "memory-plus": { "command": "uvx", "args": ["-q", "memory-plus@latest"], "env": { "GOOGLE_API_KEY": "<YOUR_API_KEY>" } } } } 4. Restart your MCP client to initialize the server.

What you can do with Memory-Plus

• Storing project-specific architecture decisions and coding conventions across sessions so multiple AI assistants retain full project context without re-prompting. - Ingesting local documentation files directly into persistent memory to allow fast semantic retrieval during active code generation sessions. - Searching past debugging solutions and error workarounds using natural language topics or keywords to accelerate troubleshooting across projects. - Visualizing stored concepts and knowledge clusters as interactive graphs to audit and explore the AI agent's internal memory map. - Tracking revisions to design decisions over time by leveraging built-in memory versioning to maintain a full history of updates.

Key facts

• Open Source
• https://github.com/Yuchen20/Memory-Plus
• AI Memory & Context, Notes & Knowledge Management
• context, local, memory, persistence, rag

Part of MCP Servers

How do I install Memory-Plus?

You can run Memory-Plus using uvx by adding it to your MCP client configuration file, such as Cursor or Cline settings. Specify uvx as the command with the arguments -q and memory-plus@latest, and provide your GOOGLE_API_KEY in the environment section.

What can Memory-Plus do?

Memory-Plus allows AI assistants to record, retrieve, update, and delete persistent memories across chat sessions. It supports file imports for context ingestion, tracks older versions when memories are modified, and provides interactive graph clusters to visualize relationship networks among stored notes.

Which MCP clients work with Memory-Plus?

Memory-Plus works with standard Model Context Protocol clients including VS Code, Cursor, Cline, and custom setups using the FastMCP framework or MCP Inspector for local testing and development.

Is Memory-Plus open source?

Yes, Memory-Plus is an open source project. It is licensed under the Apache License 2.0 and its source code is hosted publicly on GitHub.

Why does Memory-Plus take long to run the first time?

On the initial run, Memory-Plus must fetch and install several dependencies. This initial download typically takes around one minute, after which subsequent executions start much faster.