Getting Started with CrewAI + Valkey

What is CrewAI + Valkey?

CrewAI lets you build teams of AI agents that collaborate on tasks. The problem: agents forget everything between runs. Valkey gives them a shared memory that persists across executions, so your code review agent remembers past vulnerabilities and your research agent builds on previous findings.

CrewAI provides the agent orchestration, Valkey provides the memory backend:

• Sub-millisecond reads - memory recall in ~0.1ms via GLIDE
• Vector search - FT.SEARCH with HNSW for semantic recall
• JSON storage - JSON.SET stores structured memory records natively
• TTL - memories auto-expire with EXPIRE

Step 1: Start Valkey

Docker installed and Python 3.10+ required. AWS credentials needed for Bedrock (later guides).

docker run -d --name valkey -p 6379:6379 valkey/valkey-bundle:latest

The valkey-bundle image includes JSON and Search modules. Verify:

docker exec valkey valkey-cli PING
# PONG

Step 2: Install Dependencies

pip install crewai valkey-glide boto3

Why GLIDE? Valkey GLIDE is the official Valkey client with a Rust core for high performance. It supports both standalone Valkey and ElastiCache for Valkey clusters. Unlike redis-py, GLIDE is purpose-built for Valkey and supports all Valkey-specific features including FT.* search commands.

Step 3: Connection Configuration

from dataclasses import dataclass
import os

@dataclass(frozen=True)
class ValkeyConfig:
    host: str
    port: int
    use_tls: bool
    password: str | None

def get_valkey_config() -> ValkeyConfig:
    return ValkeyConfig(
        host=os.environ.get("VALKEY_HOST", "localhost"),
        port=int(os.environ.get("VALKEY_PORT", "6379")),
        use_tls=os.environ.get("VALKEY_TLS", "false").lower() in ("true", "1"),
        password=os.environ.get("VALKEY_PASSWORD") or None,
    )

Step 4: Connect and Test

import asyncio
from glide import GlideClient, GlideClientConfiguration, NodeAddress

async def test_connection():
    config = get_valkey_config()
    glide_config = GlideClientConfiguration(
        [NodeAddress(config.host, config.port)],
        use_tls=config.use_tls,
    )
    client = await GlideClient.create(glide_config)

    # Test basic operations
    await client.set("test:hello", "world")
    val = await client.get("test:hello")
    print(f"✅ Connected! Got: {val.decode()}")

    # Clean up
    await client.delete(["test:hello"])

asyncio.run(test_connection())
# ✅ Connected! Got: world

How It Works Under the Hood

Connection works. Next, we'll build the ValkeyStorage backend that implements CrewAI's storage protocol.

Next: 02 Memory Storage Backend →