Eternity

Eternity MCP Server provides a private, local memory layer for LLMs and AI agents. It stores and retrieves memories across sessions using semantic search over text and PDFs, enabling agents to remember context, conversations, and documents without leaving your machine.

How to use

You connect an MCP-compatible client to the Eternity MCP Server to store and retrieve memories. Use it as your agent's long-term memory: before answering, recall relevant context from Eternity, generate a response with that context, and then memorize the valuable bits of the interaction back into Eternity. You can persist simple memories, multi-file inputs (including PDFs), and chat histories, then search by meaning rather than just keywords.

How to install

Prerequisites: you should have Git installed and a Python 3.9+ environment available. You may also need a terminal/command prompt with internet access.

git clone https://github.com/danttis/eternity-mcp.git
cd eternity

# Optional: create and activate a virtual environment
python3 -m venv venv
source venv/bin/activate  # on Windows use venv\Scripts\activate

# Install dependencies (adjust if your environment uses a different package manager)
pip install -r requirements.txt

# Start the server
eternity

Additional sections

Configuration and running notes: Eternity runs locally and exposes a memory API and a simple web UI. By default, the server is available at http://localhost:8000. You can feed it plain text memories or PDFs, and you can perform semantic searches to retrieve context based on meaning. To integrate with LangGraph or LangChain, connect your agent’s memory layer to Eternity’s API and follow the same recall/generate/memorize pattern used in typical long-term memory workflows.

Security and privacy: all data remains on your machine. No memory data needs to be sent to external services unless you explicitly configure such behavior in your client workflows.

Notes: if you plan to ingest large PDFs or multi-session histories, consider sizing your local vector store and ensuring your machine has sufficient disk space and memory for smooth operation.

Available tools

add

Store a memory by sending content and optional tags to the server to be indexed and persisted for future retrieval.

search

Query the stored memories using natural language to retrieve context relevant to the current task.

ui

Web UI endpoint to view recent memories and interact with the local memory store.