AG2 (AutoGen) + Memanto

Persistent long-term memory for AG2 agents via Memanto — remember → recall → answer tools your assistant can call during chat.

Status: Preview package in this folder. Run unit tests before relying on it in production.

Features

• Drop-in tools — register_memanto_tools() registers remember, recall, and answer in one call
• AG2-native — Plain functions with Annotated type hints for @register_for_llm / @register_for_execution
• GroupChat — Register the same agent_id on multiple assistants; one executor runs tool calls
• Selective tools — include_remember, include_recall, include_answer
• Shared namespace — One Memanto agent_id per team or app

Installation (preview)

From the Memanto repo:

pip install -e .                      # repo root → memanto
pip install -e integrations/ag2       # memanto-ag2
pip install "memanto-ag2[ag2]"        # optional: AG2 runtime for examples

Package name: install ag2 (AI agents), not a2g (unrelated bioinformatics tool).
Version: use ag2>=0.9,<1 — ag2 1.x is a new framework; this integration uses from autogen import AssistantAgent (0.9.x).

Set your Moorcheh API key:

export MOORCHEH_API_KEY=your_key

Quick start

import os
from autogen import AssistantAgent, UserProxyAgent
from memanto.cli.client.sdk_client import SdkClient
from memanto_ag2 import openai_llm_config, register_memanto_tools

llm_config = openai_llm_config("gpt-4o-mini")

assistant = AssistantAgent(
    name="assistant",
    llm_config=llm_config,
    system_message="You are a helpful assistant with long-term memory.",
)
user_proxy = UserProxyAgent(name="user", human_input_mode="NEVER")

client = SdkClient(api_key=os.environ["MOORCHEH_API_KEY"])
register_memanto_tools(
    assistant,
    executor=user_proxy,
    client=client,
    agent_id="my-ag2-team",
)

user_proxy.initiate_chat(
    assistant,
    message="Remember that I prefer Python over JavaScript.",
)

GroupChat with shared memory

from autogen import AssistantAgent, UserProxyAgent, GroupChat, GroupChatManager, LLMConfig
from memanto_ag2 import register_memanto_tools

# ... LLMConfig + agents ...

executor = UserProxyAgent(name="executor", human_input_mode="NEVER")
for agent in [researcher, writer]:
    register_memanto_tools(agent, executor=executor, client=client, agent_id="team-memory")

group_chat = GroupChat(agents=[researcher, writer, executor], messages=[])
manager = GroupChatManager(groupchat=group_chat)

Tool mapping

Configuration

from memanto_ag2 import configure, register_memanto_tools

configure(source="ag2-prod", recall_limit=15)

register_memanto_tools(assistant, executor=executor, agent_id="my-bank")

Tests

Unit tests (no API key)

From the repository root (after pip install -e . and pip install -e integrations/ag2):

python -m pytest integrations/ag2/tests/test_tools.py -q

Live smoke script (real API)

From the repository root (no PYTHONPATH needed):

export MOORCHEH_API_KEY=your-moorcheh-api-key
python integrations/ag2/scripts/smoke_test.py

Optional Step 6 runs a real AG2 chat when OPENAI_API_KEY is set and ag2 is installed.

Expected: steps 1–5 pass, ending with PASS — AG2 Memanto smoke test completed. Exit code 0.

Pytest live smoke

From the repository root:

export MOORCHEH_API_KEY=your-moorcheh-api-key
python -m pytest integrations/ag2/tests/test_tools_live.py -q

Requirements

• Python 3.10+
• memanto with valid MOORCHEH_API_KEY
• AG2 >= 0.9.0 for runtime examples (pip install ag2)

Related

• LangGraph integration
• AgentCore integration
• Agent Integration Guide