A Mem MCP

🚀 A-MEM: Self-evolving memory for coding agents

A-MEM is a self-evolving memory system designed for coding agents. Different from simple vector stores, it automatically organizes knowledge into a Zettelkasten-style graph with dynamic relationships. Memories not only get stored but also evolve and connect over time. Currently, it has been tested with Claude Code, and support for other MCP-compatible agents is in the pipeline.

mcp-name: io.github.DiaaAj/a-mem-mcp

🚀 Quick Start

📦 Installation

pip install a-mem

💻 Adding to Claude Code

claude mcp add a-mem -s user -- a-mem-mcp \
  -e LLM_BACKEND=openai \
  -e LLM_MODEL=gpt-4o-mini \
  -e OPENAI_API_KEY=sk-...

That's all! A session-start hook will be installed automatically to remind Claude to use the memory.

⚠️ Important Note

Memory is stored per-project in ./chroma_db. For global memory across all projects, see Memory Scope.

🗑️ Uninstallation

a-mem-uninstall-hook   # Remove hooks first
pip uninstall a-mem

🔧 How It Works

t=0              t=1                t=2

                 ◉───◉             ◉───◉
 ◉               │                 ╱ │ ╲
                 ◉                ◉──┼──◉
                                     │
                                     ◉

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━▶
            self-evolving memory

1. Add a memory → A-MEM extracts keywords, context, and tags via LLM.
2. Find neighbors → Searches for semantically similar existing memories.
3. Evolve → Decides whether to link, strengthen connections, or update related memories.
4. Store → Persists to ChromaDB with full metadata and relationships.

The outcome is a knowledge graph that becomes smarter over time, not just larger.

✨ Features

Self-Evolving Memory

Memories are not static. When new knowledge is added, A-MEM automatically finds related memories, strengthens connections, updates context, and evolves tags.

Semantic + Structural Search

Combines vector similarity with graph traversal. You can find memories by meaning and then explore their connections.

Peek and Drill

Start with breadth-first search to capture relevant memories via lightweight metadata (id, context, keywords, tags). Then, drill depth-first into specific memories with read_memory_note for full content. This approach minimizes token usage while maximizing recall.

🛠️ MCP Tools

A-MEM provides 8 tools for your coding agent:

💻 Usage Examples

# The agent calls these automatically, but here's what happens:

# Store a memory (returns task_id immediately)
add_memory_note(content="Auth uses JWT in httpOnly cookies, validated by AuthMiddleware")

# Search later
search_memories(query="authentication flow", k=5)

# Deep search with connections
search_memories_agentic(query="security", k=5)

⚙️ Advanced Configuration

JSON Config

For more control, edit ~/.claude/settings.json (global) or .claude/settings.local.json (project):

{
  "mcpServers": {
    "a-mem": {
      "command": "a-mem-mcp",
      "env": {
        "LLM_BACKEND": "openai",
        "LLM_MODEL": "gpt-4o-mini",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Environment Variables

Memory Scope

• Project-specific (default): Each project has isolated memory in ./chroma_db.
• Global: Share across projects by setting CHROMA_DB_PATH=~/.local/share/a-mem/chroma_db.

Alternative Backends

Ollama (local, free)

claude mcp add a-mem -s user -- a-mem-mcp \
  -e LLM_BACKEND=ollama \
  -e LLM_MODEL=llama2

OpenRouter (100+ models)

claude mcp add a-mem -s user -- a-mem-mcp \
  -e LLM_BACKEND=openrouter \
  -e LLM_MODEL=anthropic/claude-3.5-sonnet \
  -e OPENROUTER_API_KEY=sk-or-...

Hook Management (Claude Code)

The session-start hook reminds Claude to use memory tools. It is installed automatically with Claude Code, but you can manage it manually:

a-mem-install-hook     # Install/reinstall hook
a-mem-uninstall-hook   # Remove hook completely

🐍 Python API

You can use A-MEM directly in Python (works with any agent or application):

from agentic_memory.memory_system import AgenticMemorySystem

memory = AgenticMemorySystem(
    llm_backend="openai",
    llm_model="gpt-4o-mini"
)

# Add (auto-generates keywords, tags, context)
memory_id = memory.add_note("FastAPI app uses dependency injection for DB sessions")

# Search
results = memory.search("database patterns", k=5)

# Read full details
note = memory.read(memory_id)
print(note.keywords, note.tags, note.links)

📚 Research

A-MEM implements concepts from the paper:

A-MEM: Agentic Memory for LLM Agents Xu et al., 2025 arXiv:2502.12110