mcp-external-memory

An MCP server that gives LLMs persistent, searchable semantic memory.

Install

pip install mcp-external-memory

Usage

from mcp_external_memory import memory_store, memory_search

# Store a memory
result = memory_store(content="Alice prefers dark mode", namespace="users", tags=["alice", "ui"])

# Search memories
results = memory_search(query="what does Alice prefer?", namespace="users")

CLI

mcp-external-memory --help

API

Tools

| Tool | Description | |------|-------------| | memory_store | Persist text + optional namespace/tags/metadata | | memory_search | Semantic search (cosine similarity) over all memories | | memory_get | Retrieve a single memory by ID | | memory_delete | Delete a memory by ID | | memory_list | List memories with optional namespace/tag filter + pagination | | memory_stats | Count of memories, namespaces, DB path | | memory_update | Update an existing memory |

Embedding Backends

The server supports multiple embedding backends:

• TF-IDF (default): Pure Python, no external dependencies
• OpenAI: Uses text-embedding-3-small model
• Ollama: Local embeddings with Ollama

Set via MEMORY_EMBED_BACKEND environment variable.

Development

git clone https://github.com/daedalus/mcp-external-memory.git
cd mcp-external-memory
pip install -e ".[test]"

# run tests
pytest

# format
ruff format src/ tests/

# lint
ruff check src/ tests/

# type check
mypy src/

MCP Registry

mcp-name: io.github.daedalus/mcp-external-memory