Home › Other › a-mem-mcp

a-mem-mcp

Self-evolving memory system for AI agents

Python

★ 35 stars Last pushed 2026-01-17 License: MIT

View repository →

Connect

Review any command before running it. Package names and URLs come from the server's own registry entry.

Package (pypi 0.2.1)

uvx a-mem

Or add to your MCP client config:

{
  "mcpServers": {
    "a-mem-mcp": {
      "command": "uvx",
      "args": [
        "a-mem"
      ],
      "env": {
        "LLM_BACKEND": "<YOUR_VALUE>",
        "LLM_MODEL": "<YOUR_VALUE>",
        "OPENAI_API_KEY": "<YOUR_VALUE>",
        "EMBEDDING_MODEL": "<YOUR_VALUE>",
        "CHROMA_DB_PATH": "<YOUR_VALUE>"
      }
    }
  }
}

LLM_BACKEND — LLM backend to use (openai, ollama, sglang, or openrouter)
LLM_MODEL — LLM model name (e.g., gpt-4o-mini, llama2, etc.)
OPENAI_API_KEY required secret — OpenAI API key (required if LLM_BACKEND=openai)
EMBEDDING_MODEL — Sentence transformer model for embeddings
CHROMA_DB_PATH — ChromaDB storage directory path