Memoria

Memoria is an MCP server that runs entirely on your machine to help your AI understand code dependencies more accurately. By analyzing your local git history and file changes, it reveals files that tend to change together and surfaces hidden dependencies that can cause regressions when you refactor or adjust code.

How to use

You integrate Memoria with your MCP client so your AI can run forensic analyses before editing code. Use Memoria to run a full or quick analysis of a target file, see which files are tightly coupled, assess historical risk, and identify stale dependencies. Your AI can then decide to update all related files together, reducing crashes and regressions.

How to install

Prerequisites: ensure you have Node.js 18 or newer installed and access to a Git repository with history.

Choose one of the supported installation methods shown below and follow the exact commands.

Config example for MCP client

Configure Memoria in your MCP client to run the local, self-contained server. The following is a standard local setup using npx to launch Memoria.

{
  "mcpServers": {
    "memoria": {
      "command": "npx",
      "args": ["-y", "@byronwade/memoria"]
    }
  }
}

Usage notes

After configuring Memoria, restart your AI tool to load the MCP server. You can then ask your AI to analyze files before editing, or invoke the full CLI to perform forensics, risk assessments, or to list coupled files. Memoria provides a risk score, a list of coupled files, detected stale dependencies, and evidence from code diffs.

CLI commands and outputs

Memoria ships a CLI that mirrors the AI’s capabilities. Use it to run full forensics, quick risk checks, or to discover which files import or depend on a target.

Configuration and security

Memoria runs 100% locally. It does not upload code or require API keys. It analyzes your local .git folder and respects your project boundaries, so your data never leaves your machine.

Troubleshooting

If you see a “Tool not found” message or the analyze_file capability isn’t available, restart your AI tool and verify that the MCP server loads on startup. Ensure your JSON config is valid and that you’re in a git repository with history.

Available tools

analyze_file

Forensic analysis of a target file returning coupled files, risk score, stale dependencies, and evidence from code diffs.

risk

Quick risk assessment of a target file or path to gauge bug-prone areas and historical volatility.

coupled

Show files that frequently change together with a given file to reveal implicit dependencies.

importers

Find files that import or depend on the target file.

history

Search git history to understand why code was written and how changes evolved over time.