Python SDK — neo4j-agent-memory

The Python SDK is the original implementation of the agent-memory contract. It can talk to Neo4j directly over Bolt or to the hosted NAMS service over HTTP.

Install

pip install neo4j-agent-memory
# or, with uv:
uv add neo4j-agent-memory

Optional extras (extractors, framework integrations, embeddings):

pip install "neo4j-agent-memory[all]"

Quick start

from neo4j_agent_memory import MemoryClient, MemorySettings

settings = MemorySettings(
    neo4j={"uri": "bolt://localhost:7687", "password": "password"}
)

async with MemoryClient(settings) as client:
    msg = await client.short_term.add_message("session-1", "user", "Hello")
    await client.long_term.add_entity("Alice", "PERSON")
    context = await client.get_context("hello")

For the hosted service, swap the neo4j={…} config for nams={"api_key": "nams_…"} — see NAMS Quickstart.

Where to go next

Tutorial: First agent memory
Tutorial: Add memory to a chatbot
How-To Guides — entity extraction, batch processing, framework integrations, deduplication, and more.
Configuration reference
CLI reference

Source and releases

Package: neo4j-agent-memory on PyPI
Source: github.com/neo4j-labs/agent-memory
Release tags: python-v* (e.g. python-v0.4.1)
Changelog: CHANGELOG.md

Looking for the TypeScript SDK?

See TypeScript SDK. The two SDKs share the same memory model and back onto the same NAMS service, so a Python and a TypeScript agent can read and write the same memory.