memory-bank

Memory Bank MCP Server exposes a remote, multi-project memory bank management service via MCP. It lets you access memory bank files stored per project, perform read/write/update operations, list available projects, and enumerate files within each project, all through the MCP protocol with type-safe operations and proper error handling. The server is designed to isolate project memory banks to ensure secure multi-tenant usage and provide a centralized interface for remote clients to manage memory banks. Typical usage involves configuring the MCP client (Claude, Cursor, Roo Code, or other MCP clients) with the server command and environment details, including the root directory where memory banks live, and then issuing MCP operations such as memory_bank_read, memory_bank_write, memory_bank_update, list_projects, and list_project_files.

Prerequisites:

• Node.js (LTS) and npm installed
• Optional: Smithery CLI for automatic configuration

Install and run the MCP server client locally:

1. Install or run via npx (no local install required):

# Quick install/run using npx (no local install needed)
npx -y @allpepper/memory-bank-mcp

2. If you prefer a manual install locally (for development):

# Clone or download the repository
# Install dependencies
npm install

# Build (if applicable) and run
npm run build
npm run dev   # or npm start depending on package.json

3. If using Smithery for automatic configuration:

npx -y @smithery/cli install @allpepper/memory-bank-mcp --client claude

Prerequisites summary:

• Node.js and npm
• Access to a memory bank root directory to store per-project banks
• Optional: Docker for containerized runs as shown in the documentation

Notes and tips:

• The MEMORY_BANK_ROOT environment variable must point to the base directory where per-project memory banks are stored. Ensure proper filesystem permissions.
• When integrating with MCP clients (e.g., Claude, Cursor, Roo Code), include the provided environment and arguments in the client's MCP settings as demonstrated in the README.
• For Docker usage, you can mount a host directory to the container path and pass MEMORY_BANK_ROOT accordingly. Ensure you expose only the necessary paths and maintain proper isolation between projects.
• If autoApprove is enabled for certain operations, those actions will not prompt for user confirmation.
• Use the provided example configuration as a template for your MCP client settings, and adjust memory bank root paths and project scopes as needed.
• When running in production, consider additional security hardening, such as access controls, network isolation, and secure storage for credentials if any are used by your MCP workflow.