agentmemo

Persistent memory for AI agents. Store, semantically retrieve, and recall context across sessions. Pure standard library (no required dependencies).

pip install agentmemo

Quick start

from agentmemo import MemoryClient

# Get a free key (or pass an existing one)
key = MemoryClient.signup("my-agent")["api_key"]
mem = MemoryClient(key)

# Store
mem.store(
    user_id="user_123",
    agent_id="assistant",
    content="User prefers dark mode and works in TypeScript.",
    tags=["preference"],
    importance=8,
)

# Retrieve semantically
hits = mem.search(user_id="user_123", query="what are the user's preferences?")

# Context for an LLM system prompt
ctx = mem.context(user_id="user_123", format="anthropic")["context"]

API

• store(user_id, agent_id, content, metadata=None, ttl_seconds=None, tags=None, namespace="default", importance=5, outcome="unknown", detect_conflicts=False)
• search(user_id, query, agent_id=None, limit=10, namespace=None, tags=None, min_importance=None)
• delete(id=None, user_id=None, agent_id=None)
• context(user_id, agent_id=None, max_tokens=2000, format="raw")
• feedback(memory_id, outcome, confidence=1.0)
• batch(memories)
• stats(user_id=None), usage()
• MemoryClient.signup(name) — static, returns a free API key
• Async: await mem.async_store(...), await mem.async_search(...) (requires pip install agentmemo[async])

Errors raise AgentMemoError with .status, .code, .body.

Docs: https://agentmemo.dev/docs · Apache-2.0 · Built by Dr. Nadeem Shaikh