claude-self-reflect

Claude Self-Reflect is a local-first memory augmentation tool for Claude that stores and retrieves conversation context to provide persistent recall across sessions. This MCP server wraps the claude-self-reflect package, enabling you to run the memory-augmented Claude workflow locally or in a containerized environment. It supports both fully local embeddings and optional cloud embeddings via Voyage AI, allowing you to choose between private on-device search and higher-accuracy cloud-powered search. Once running, you can interact with Claude as usual, while the plugin automatically indexes conversations, enabling rapid search and retrieval of past discussions to preserve context and improve continuity.

Key capabilities include: quick installation and setup with a single global CLI, automatic Docker-based deployment if you prefer containerization, and automatic migration between versions. The tool emphasizes privacy and local data by default, with cloud mode available for enhanced search accuracy when you supply a Voyage AI key. You can manage memory across projects through Ralph Loop Memory integrations when using the related ecosystem tooling, enabling cross-project memory sharing while keeping data accessible locally when desired.

Prerequisites:

• Node.js (v14+ recommended) and npm installed on your system
• Optional: Docker installed if you want Docker-based deployment

Steps:

1. Install the claude-self-reflect CLI globally

npm install -g claude-self-reflect

2. Run the quick setup (optional but recommended)

claude-self-reflect setup

3. If you plan to use Cloud Mode with Voyage AI, obtain a Voyage AI key and configure it when prompted or via the CLI

claude-self-reflect setup --voyage-key=YOUR_ACTUAL_KEY_HERE

4. Start or verify the MCP integration in your environment (the CLI will handle Docker or local setup as configured during setup)

claude-self-reflect start

Prerequisites recap:

• A working Node.js/npm environment
• Optional Voyage AI credentials for cloud-based embeddings
• Docker if you want containerized deployment (the setup can run in Docker by default)

Tips and common issues:

• If you run into permissions errors installing the CLI globally, consider using a Node version manager (nvm) and ensuring npm global bin is in your PATH.
• Cloud Mode requires a Voyage AI key; without it, memory remains local with 384-dimension embeddings for searches.
• Data stays local by default; review Docker storage settings if you run out of space.
• When upgrading claude-self-reflect, use npm update -g claude-self-reflect to auto-migrate; the CLI notes migrations for breaking changes.
• If using Ralph Loop Memory integrations, ensure compatible versions of the related plugins and hooks are installed as per the Ralph memory docs.
• For troubleshooting, consult the CLI’s built-in help: claude-self-reflect --help