In Memoria

In Memoria is a local MCP server that gives AI assistants persistent memory of a codebase across sessions.

What you can do

• Provides persistent codebase intelligence so an assistant can recall a project's structure and context between sessions.
• Keeps memory across sessions rather than rebuilding context each time.
• Can use OpenAI embeddings when a key is supplied, otherwise falls back to local transformers.js embeddings.
• Runs as a local stdio process on the user's machine; no tool list is published.

When to use it

Useful when you work with an assistant on the same codebase over many sessions and want it to retain project context instead of re-reading everything each time.

Requirements

Runs locally as an npm package (in-memoria) over stdio, so Node.js and a desktop MCP client are needed. Optionally set the OPENAI_API_KEY environment variable for OpenAI embeddings; without it, local transformers.js embeddings are used. No authentication is declared.