This plugin provides memory persistent, maintain key decisions, safety and rules across sessions. No more explanations of your configuration, or concerns that the agent will perform a dangerous operation such as force-pushing to your main branch.

Here is what it does:

It stores information about your technical infrastructure, design patterns, choices and notes and retrieves this information as needed. Prevents execution of unsafe commands, not just by polite request.

Retrieves and stores information missed by the agent during each session. Operates silently in the background while you use Claude Code normally.

It includes 19 MCP tools, is fully open source and distributed under the MIT license.