CrewAI × CLS++ (native external-memory provider)

Plug CLS++ in as CrewAI's persistent memory layer. Facts your agents learn during one crew run are stored in CLS++ and automatically recalled — as task context — on later runs, across processes and machines. This is the same slot CrewAI documents for Mem0 (a third-party ExternalMemory storage), routed through the CLS++ brain instead.

Install

pip install clsplusplus[crewai]

crewai is an optional extra — importing clsplusplus never requires it. You only need it installed to actually run a Crew.

Configure

Grab an API key from your profile: https://www.clsplusplus.com/profile#api-keys

export CLS_API_KEY="cls_live_..."
# optional — defaults to https://www.clsplusplus.com
export CLS_BASE_URL="https://www.clsplusplus.com"

The provider reads CLS_API_KEY / CLS_BASE_URL from the environment (the same defaults as the rest of the CLS++ SDK). You can also pass api_key= / url= to CLSMemoryStorage(...) explicitly.

Wire it into a Crew

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

from clsplusplus.integrations.crewai import CLSMemoryStorage

# `user` is the CLS++ namespace ("whose brain"). Use a per-crew or per-tenant
# id so memory is scoped correctly.
cls_storage = CLSMemoryStorage(user="support-crew")

researcher = Agent(
    role="Researcher",
    goal="Answer questions using what the crew has learned before",
    backstory="You remember everything across runs via CLS++.",
)

task = Task(
    description="What contact preferences do we know about this customer?",
    expected_output="A short summary of known preferences.",
    agent=researcher,
)

crew = Crew(
    agents=[researcher],
    tasks=[task],
    memory=True,
    external_memory=ExternalMemory(storage=cls_storage),
)

result = crew.kickoff()
print(result)

On each kickoff(), CrewAI:

Recalls — calls cls_storage.search(task_description) and injects the returned memories into the agent's task context.
Stores — calls cls_storage.save(value, metadata) with what the agents produced, persisting it to CLS++ for the next run.

What the provider implements

CLSMemoryStorage implements CrewAI's storage contract (crewai.memory.storage.interface.Storage):

Method	CLS++ behavior
`save(value, metadata)`	Stores the text in CLS++ (`Brain.learn`), forwarding CrewAI metadata.
`search(query, limit, score_threshold)`	Semantic recall (`Brain.ask`); returns `list[dict]` where each item has a `"content"` key (what CrewAI's `ContextualMemory` reads).
`reset()`	Documented no-op — CLS++ is durable shared memory; manage retention via the CLS++ API or `Brain.forget`. Override the method if you want `crew.reset_memories()` to clear the namespace.

Notes