CrewAI Integration

Use SynapseCrewStorage to give your CrewAI crews persistent, encrypted memory. Drop-in replacement for CrewAI's default storage — every memory passes through PII sanitization and AES-256-GCM encryption.

Overview

CrewAI is a framework for orchestrating autonomous AI agents into collaborative crews. By default, CrewAI stores agent memory in local storage. SynapseCrewStorage replaces this with Synapse Layer's Cognitive Security Pipeline — giving every memory automatic PII redaction, intent classification, and AES-256-GCM encryption at rest.

The integration implements CrewAI's Storage interface, so existing crews work without code changes beyond swapping the storage backend.

Installation

bash
pip install synapse-layer crewai

Quick Start

Standalone example — no LLM required:

python
from synapse_memory.integrations.crewai_memory import SynapseCrewStorage
from crewai.memory.types import MemoryRecord

# Initialize Synapse Layer as CrewAI's storage backend
storage = SynapseCrewStorage(agent_id="research-crew")

# Store memories through the Cognitive Security Pipeline
records = [
    MemoryRecord(
        content="User prefers concise technical reports.",
        scope="/crew/research",
        categories=["preference"],
        importance=0.8,
    ),
    MemoryRecord(
        content="Project deadline is next Friday.",
        scope="/crew/research",
        categories=["context"],
        importance=0.9,
    ),
]

for record in records:
    storage.save(record)

print(f"Stored {len(records)} memories.")

# Search across stored memories
results = storage.search("What are the project deadlines?", top_k=3)
for r in results:
    print(f"  [{r.score:.2f}] {r.content}")

Crew Integration

Use with a full CrewAI crew:

python
from crewai import Agent, Crew, Task
from crewai.memory.unified_memory import Memory
from synapse_memory.integrations.crewai_memory import SynapseCrewStorage

# Synapse Layer handles persistence and encryption
storage = SynapseCrewStorage(agent_id="research-crew")
memory = Memory(storage=storage)

researcher = Agent(
    role="Senior Researcher",
    goal="Find and summarize key insights",
    backstory="Expert analyst with deep domain knowledge.",
)

task = Task(
    description="Research the latest trends in AI agent memory.",
    expected_output="A summary of key trends.",
    agent=researcher,
)

crew = Crew(
    agents=[researcher],
    tasks=[task],
    memory=memory,  # Synapse Layer handles all persistence
)

result = crew.kickoff()

Note

Every memory stored by the crew passes through PII sanitization, intent classification, and AES-256-GCM encryption before persistence — automatically.

API Reference

SynapseCrewStorage

MethodDescription
save(record)Store a MemoryRecord through the security pipeline
search(query, top_k)Semantic search across stored memories
reset()Clear all memories for this agent

Source: synapse_memory/integrations/crewai_memory.py