Quickstart¶

Get an ADK agent running with Redis-backed sessions and long-term memory in three steps. For the concepts behind each feature, see the Concepts section.

See also: ADK + Redis on redis.io for additional coverage.

1. Start infrastructure¶

Provide Redis Agent Memory connection settings:

export REDIS_MEMORY_BACKEND="redis-agent-memory"
export REDIS_AGENT_MEMORY_API_BASE_URL="https://..."
export REDIS_AGENT_MEMORY_STORE_ID="..."
export REDIS_AGENT_MEMORY_API_KEY="..."

These are the same variable names used by the managed memory quickstart and the integration tests. For the open source self-hosted Agent Memory Server, use REDIS_MEMORY_BACKEND="opensource-agent-memory", point REDIS_AGENT_MEMORY_API_BASE_URL at your server, and omit the API key and store ID unless your deployment requires them.

2. Install dependencies¶

pip install google-adk "adk-redis[memory]"

3. Wire services into an agent¶

import os

from google.adk import Agent
from google.adk.agents.callback_context import CallbackContext
from google.adk.runners import Runner
from google.adk.tools import load_memory
from google.adk.tools import preload_memory
from adk_redis import (
    REDIS_AGENT_MEMORY_BACKEND,
    RedisSessionMemoryService,
    RedisSessionMemoryServiceConfig,
    RedisLongTermMemoryService,
    RedisLongTermMemoryServiceConfig,
)

# REDIS_AGENT_MEMORY_BACKEND and OPENSOURCE_AGENT_MEMORY_BACKEND are typo-safe
# aliases for the "redis-agent-memory" and "opensource-agent-memory" strings.
backend = os.getenv("REDIS_MEMORY_BACKEND", REDIS_AGENT_MEMORY_BACKEND)

session_service = RedisSessionMemoryService(
    config=RedisSessionMemoryServiceConfig(
        backend=backend,
        api_base_url=os.environ["REDIS_AGENT_MEMORY_API_BASE_URL"],
        api_key=os.environ.get("REDIS_AGENT_MEMORY_API_KEY"),
        store_id=os.environ.get("REDIS_AGENT_MEMORY_STORE_ID"),
        default_namespace="my_app",
    )
)

memory_service = RedisLongTermMemoryService(
    config=RedisLongTermMemoryServiceConfig(
        backend=backend,
        api_base_url=os.environ["REDIS_AGENT_MEMORY_API_BASE_URL"],
        api_key=os.environ.get("REDIS_AGENT_MEMORY_API_KEY"),
        store_id=os.environ.get("REDIS_AGENT_MEMORY_STORE_ID"),
        default_namespace="my_app",
    )
)


async def after_agent(callback_context: CallbackContext):
    await callback_context.add_session_to_memory()


agent = Agent(
    name="memory_agent",
    model="gemini-2.0-flash",
    instruction="You are a helpful assistant with long-term memory.",
    tools=[preload_memory, load_memory],
    after_agent_callback=after_agent,
)

runner = Runner(
    agent=agent,
    app_name="my_app",
    session_service=session_service,
    memory_service=memory_service,
)

Launch the agent with the ADK runtime (adk web):

adk web my_app

Try it out:

"Hi, I'm Alice. I love pizza and Python."
Start a new session: "What do you remember about me?"

What next?¶

Goal	Page
Understand sessions, memory, search, and caching	Concepts
Configure session or memory services in detail	Session service how-to, Memory service how-to
Give the LLM explicit memory tools	Sessions + Memory MCP + Tools
Add vector or hybrid search	Search tools how-to
Reduce LLM cost with semantic caching	Semantic Caching
See a full working example	Examples
Run or deploy your agent	ADK runtime