memory-bank

The Memory Bank MCP Server provides a persistent, structured vocabulary for project knowledge. It stores global memory and branch-specific memories in a JSON format, enabling AI agents to maintain context across sessions and tasks. The server exposes MCP tools for creating, querying, and updating memory documents, and it coordinates a lifecycle of memory banks to support consistent language model reasoning over time. Once running, you can interact with it to add, retrieve, or modify memory entries, and to run memory-related tooling within your agent workflows.

Prerequisites:

• Node.js (as specified by the project and .tool-versions)
• Yarn (v1.x)

Installation steps:

1. Clone the repository: git clone https://github.com/t3ta/memory-bank-mcp-server.git
2. Navigate into the project root: cd memory-bank-mcp-server
3. Install dependencies (workspaces): yarn install
4. Build or prepare packages if required by the project (per package README): yarn build
5. Start the MCP server (from the monorepo root): yarn workspace @memory-bank/mcp start --docs /path/to/your/docs

Notes:

• The exact start command may vary by package configuration; consult packages/mcp/README.md for server options and flags.
• Ensure the docs path exists and is writable by the server.

Tips and caveats:

• The server stores memory in JSON format; ensure backups and proper access controls if running in shared environments.
• If you upgrade packages, re-run yarn install to refresh workspace symlinks.
• Environment variable DOCS_PATH configures where the memory banks are stored for global and branch scopes; adjust as needed for your workspace.
• If you encounter port or binding issues, check for existing processes using the expected port and update configuration accordingly.