Give a CrewAI crew persistent memory. Shared across the crew.

CrewAI orchestrates multiple agents working toward a goal. By default a crew’s memory is local to a run; this tutorial wires OctaMemin as the crew’s external memory so knowledge persists across runs and is shared between agents.

Prerequisites

• Python 3.9+ and a model provider configured for CrewAI (e.g. OpenAI).
• A free OctaMem API key from platform.octamem.com.

Step 1: Install

pip install octamem 'octamem[crewai]' crewai

Set OCTAMEM_API_KEY in your environment so OctaMemStorage can authenticate without hard-coding the key.

Step 2: Back the crew with OctaMem

Pass ExternalMemory(storage=OctaMemStorage()) as the external_memoryargument when you build your crew. That’s the whole integration — CrewAI handles the read/write calls.

from crewai import Agent, Task, Crew
from crewai.memory.external.external_memory import ExternalMemory
from octamem.integrations.crewai import OctaMemStorage

researcher = Agent(
    role="Researcher",
    goal="Gather and remember key facts about the account",
    backstory="You track everything the team learns about each customer.",
)

task = Task(
    description="Summarize what we know about customer #4521.",
    expected_output="A short brief citing prior interactions.",
    agent=researcher,
)

crew = Crew(
    agents=[researcher],
    tasks=[task],
    # Back the crew's external memory with OctaMem.
    external_memory=ExternalMemory(storage=OctaMemStorage()),
)

result = crew.kickoff()
print(result)

How it works

Each memory the crew stores becomes a typed record with provenance in your OctaMem account, retrievable by natural-language query. Because it’s external and account-scoped, the same memory is available on the next run, to other agents, and even to MCP clients like Cursor and Claude. For a deeper comparison of approaches, see Mem0 vs Zep vs OctaMem.

FAQ