Skip to directory

Memory-Plus

Yuchen20 (community maintainer)

stdio8+ documented tools1

About Memory-Plus

Community-maintained local RAG memory MCP that records, retrieves, updates, visualizes, imports, and deletes persistent memories for an agent.

What you can use it for

  • persistent local memory
  • semantic retrieval
  • memory visualization
  • document import
Connection and setup
Hosting
Local process
Transport
stdio
Endpoint
Local process; no hosted endpoint documented
Authentication
Google API key for the Gemini Embedding API via GOOGLE_API_KEY.

Requirements

  • UV runtime; the repository does not state a minimum UV or Python version.
  • Google AI Studio API key set as GOOGLE_API_KEY.
  • Local filesystem access for its memory database and imported files.

Setup steps

  1. Open the official Google AI Studio API Keys page at https://aistudio.google.com/api-keys, select Create API key, and copy the generated key. If your existing Cloud project is not listed, use Dashboard > Projects > Import projects. Store it in the MCP client's secure local environment as GOOGLE_API_KEY; do not paste it into chat.
  2. Install UV using the official Astral installer or the README's pip install uv option.
  3. Configure the MCP client for stdio with command uvx, args ["-q", "memory-plus@latest"], and GOOGLE_API_KEY set through the client's local secret/environment mechanism.
  4. Restart the MCP client and verify the server appears in its MCP tool list. Do not use record, update, import_file, set_whether_to_annonimize, or delete as a smoke test because those can write data. Use tools/list only until the owner has configured an empty test profile and explicitly approves a harmless read-only test.
  5. If Create API key is unavailable, the linked Google guide says the Google Cloud project administrator may need to grant the documented project/key creation permissions; do not guess or widen permissions.
Tools 8+

Eight tool registrations were inspected in the linked source file; runtime tools/list was not performed. The server instructions describe automatic memory retrieval and capture behavior, so this inventory is not treated as runtime-exact.

  • set_whether_to_annonimize

    Changes the local anonymization setting by writing a file.

    Parameters: whether_to_annonimize
    01
  • record

    Stores memory content and metadata.

    Parameters: content, metadata
    02
  • retrieve

    Searches stored memories by semantic query.

    Parameters: query, top_k
    03
  • recent

    Returns recently recorded memories.

    Parameters: limit
    04
  • update

    Replaces a stored memory with new content and metadata.

    Parameters: memory_id, new_content, metadata
    05
  • visualize

    Creates a URL or embedded HTML snippet for a memory visualization.

    06
  • delete

    Deletes a memory by ID; its tool documentation requires explicit confirmation.

    Parameters: memory_id, confirmation
    07
  • import_file

    Imports a readable text file into the memory database.

    Parameters: file_path, metadata
    08

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

Limitations and notes
  • This community project's server instructions say the assistant should monitor conversations for personal context and record/update memories automatically; this may persist sensitive or personal information. Review and configure its behavior before using it with real data.
  • The integration exposes tools that write, update, import, or delete memory; these were not invoked.
  • Google API key needed for the documented Gemini Embedding API integration.
  • Google AI Studio key creation may require the related Google Cloud project to be imported and appropriate key/project permissions; exact minimum permissions are documented by Google.
Original sources

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.