Google ADK + MemorySync

Overview

Install google-adk-memorysync and pass MemorySyncMemoryService() to Runner(memory_service=...): all four BaseMemoryService methods are real — add_session_to_memory persists sessions (and raises on failure), search_memory powers ADK’s native load_memory and preload_memory tools. A context tool guarantees per-turn injection plus user-turn persistence, an after_model_callback persists assistant turns, and five structured tools give agents explicit memory. The only vendor ADK integration whose ingestion is not a no-op.

Setup status and requirements

Supported version

google-adk-memorysync 1.0.0 (PyPI)

Last setup review

2026-08-23

Permissions

A MemorySync API key with read and write scopes; the delete tool needs delete permission.

Limits

Requires google-adk >=2 <3 (Python 3.10+). Python only — TypeScript agents use the Vercel AI SDK or Mastra integrations.

Capabilities

• All four BaseMemoryService methods implemented — ingestion included
• Powers ADK’s native load_memory and preload_memory tools unchanged
• Repeated session ingestion converges via event-id idempotency seeds
• Context tool: guaranteed injection, deduped per invocation
• after_model_callback persists assistant turns as they happen
• Cross-app user memory by default; scope_to_app=True to silo
• Five never-raise agent tools with a read-only mode

Quick Start

from google.adk.agents import LlmAgent
from google.adk.runners import Runner
from google.adk.tools import preload_memory
from google_adk_memorysync import MemorySyncMemoryService

memory_service = MemorySyncMemoryService()

agent = LlmAgent(
    name="assistant",
    model="gemini-2.5-flash",
    instruction="You are a helpful assistant.",
    tools=[preload_memory],
)
runner = Runner(agent=agent, app_name="support",
                session_service=session_service,
                memory_service=memory_service)

Use Cases

Explore More