Skip to main content
The MemMachine adapter for the Google Agent Development Kit (ADK) provides a seamless implementation of the BaseMemoryService. This allows your Google GenAI agents to automatically store session events and retrieve relevant context using MemMachine’s hybrid episodic and semantic search.

​
Overview

By using the MemmachineMemoryService, your ADK agents can:
  • Auto-Archive Sessions: Automatically move completed ADK session events into MemMachine episodic memory.
  • Contextual Recall: Use the standard ADK load_memory tool to query past interactions across different sessions.
  • Scoped Search: Metadata like app_name and user_id are automatically preserved to ensure privacy and relevance.

​
Configuration

The adapter is configured via constructor arguments when initializing the service.
ParameterDescriptionDefault
api_keyRequired. Your MemMachine API Key.N/A
endpointThe MemMachine v2 REST API URL.https://api.memmachine.ai/v2
org_idYour organization identifier.None
project_idYour project identifier.None
1

Install with ADK Extra

To ensure all Google GenAI and ADK dependencies are included, install the client with the google-adk extra:
pip install "memmachine-client[google-adk]"
2

Initialize the Memory Service

Create an instance of the MemmachineMemoryService. This service acts as the bridge between Google ADK and your memory backend.
from memmachine.integrations.google_adk.memmachine_memory_service import MemmachineMemoryService

memory_service = MemmachineMemoryService(
    api_key="YOUR_API_KEY",
    endpoint="http://localhost:8080/api/v2", # Use for self-hosted
    org_id="my-org",
    project_id="my-project"
)
3

Equip the Agent

Add the standard load_memory tool to your LlmAgent. This allows the agent to decide when it needs to “look back” at past conversations to answer a query.
from google.adk.agents import LlmAgent
from google.adk.tools import load_memory

agent = LlmAgent(
    model="gemini-2.0-flash",
    name="MemoryAgent",
    instruction="Use the 'load_memory' tool if the answer is in past conversations.",
    tools=[load_memory]
)
4

Wire the Runner

Connect your agent, session service, and the MemMachine memory service into a Runner.
from google.adk.runners import Runner
from google.adk.sessions import InMemorySessionService

runner = Runner(
    agent=agent,
    app_name="my_adk_app",
    session_service=InMemorySessionService(),
    memory_service=memory_service
)

​
Memory Lifecycle

​
Persistence

To move a session from active memory into long-term storage (so it can be recalled by other sessions), call add_session_to_memory once a conversation is finished: 
# completed_session is a google.adk.sessions.session.Session object
await memory_service.add_session_to_memory(completed_session)

​
Retrieval

When the agent invokes load_memory, the adapter automatically filters results based on the app_name and user_id defined in the Runner, ensuring the agent only recalls information relevant to the current user.
Important: The ADK integration requires Python 3.10+ and the google-adk extra package. If you encounter import errors, verify that you have installed memmachine-client[google-adk].