MCP Memory System Server

External memory system to persist and reference decisions across chat sessions using MCP over SSH.

python

0

GitHub Stars

python

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "workerportfolio-mcp-memory-system": {
      "command": "python",
      "args": [
        "server.py"
      ]
    }
  }
}

You deploy the MCP Memory System to extract, store, and reference decisions, configurations, and operation rules across chat sessions. This external memory acts as your project’s shared brain, preserving decision continuity, normalizing information, and making governance transparent for long-running development.

How to use

Interact with the memory server through an MCP client to create, update, and query decision items. You will draft items with a clear lifecycle, mark them as draft or final, and supersede obsolete decisions while linking related versions. Use explicit user actions to confirm important changes, and rely on the memory store to provide consistent reference across sessions.

How to install

Prerequisites you need before starting:

Python 3.12 or later
Git for Windows (for SSH and tooling)
A Unix-like shell available in the VM or host environment

Install and run the MCP memory server locally on the Ubuntu VM side. Use the following commands in order:

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python server.py

Configuration and usage notes

The memory system uses a simple state model to track items as draft, final, or obsolete. Always attach a clear lifecycle to each decision item and link related versions to preserve traceability.

SSH stability between Windows and the Ubuntu VM is improved by using the Git for Windows SSH client and non-interactive private keys. This lowers the likelihood of connection drops during MCP communication.

Explicitly control when tools are invoked to avoid unintended activations by the LLM. Prefer human confirmation for critical saves to maintain stable behavior as context grows.

Security and deployment notes

Operate with non-interactive SSH keys and limit interactive prompts to essential confirmations. Maintain separate environments for development and production and ensure the memory store is isolated from public access.

Troubleshooting and lessons learned

If you encounter unstable tool triggers or unexpected saves, adjust the explicit triggers and validate lifecycle transitions in the server logic. Favor server-side state management over autonomous client decisions to reduce misfires.

Notes on future work

Potential improvements include stronger UX controls for save actions, stricter server-enforced workflows, and broader deployment options for MCP servers beyond the current Ubuntu VM setup.

Available tools

save_draft

Tool used to save a draft decision item with its metadata into the memory store, enabling later finalization and linkage to related items.