Skip to main content

CrewAI Integration

Installation​

pip install comply54 crewai

Basic usage​

from crewai import Agent, Task, Crew
from comply54 import NigeriaFintechCompliance
from comply54.crewai import Comply54CrewTool

compliance = NigeriaFintechCompliance()
compliance_tool = Comply54CrewTool(compliance=compliance)

compliance_officer = Agent(
    role="Compliance Officer",
    goal="Ensure all financial transactions comply with Nigerian regulations",
    backstory=(
        "You are a Nigerian financial compliance expert. Before any transaction "
        "is approved, you verify it against NDPA, CBN, BVN/NIN, and NFIU rules."
    ),
    tools=[compliance_tool],
    verbose=True,
)

check_task = Task(
    description=(
        "Check whether a transfer of ₦8,000,000 NGN to account 0123456789 "
        "is compliant for a KYC Tier 3 customer."
    ),
    agent=compliance_officer,
    expected_output="Compliance decision with reason and audit ID",
)

crew = Crew(agents=[compliance_officer], tasks=[check_task])
result = crew.kickoff()
print(result)

Multi-agent crew with compliance gating​

from crewai import Agent, Task, Crew, Process

compliance_tool = Comply54CrewTool(compliance=NigeriaFintechCompliance())

# Agent 1: Validates the transaction request
validator = Agent(
    role="Transaction Validator",
    goal="Validate transaction requests against Nigerian regulations",
    tools=[compliance_tool],
    backstory="Expert in CBN transaction controls and NFIU AML requirements",
)

# Agent 2: Executes only after compliance approval
executor = Agent(
    role="Transaction Executor",
    goal="Execute approved transactions securely",
    backstory="Responsible for executing pre-approved financial transactions",
)

validate_task = Task(
    description="Validate transfer of ₦3,000,000 for customer tier 2",
    agent=validator,
    expected_output="JSON compliance decision",
)

execute_task = Task(
    description="Execute the transfer if compliance check passed",
    agent=executor,
    context=[validate_task],   # receives validator output
    expected_output="Transaction execution confirmation or rejection reason",
)

crew = Crew(
    agents=[validator, executor],
    tasks=[validate_task, execute_task],
    process=Process.sequential,
)
crew.kickoff()

Tool output format​

The Comply54CrewTool returns a JSON string the agent can parse and reason about:

{
  "overall": "escalate",
  "blocked": true,
  "audit_id": "aud_f3a9b2c1",
  "violations": [
    {
      "pack": "nigeria/nfiu-aml",
      "regulation": "NFIU/MLPPA 2022",
      "action": "escalate",
      "messages": [
        "Transaction ≥ ₦5,000,000 — Currency Transaction Report required within 24 hours"
      ]
    }
  ]
}

Injecting context from crew state​

If your crew tracks KYC context, pass it via the tool input:

# The agent calls the tool with this structured input
tool_input = {
    "action": "transfer_funds",
    "params": {"amount": 5_000_000, "currency": "NGN"},
    "context": {"kyc_tier": 3, "pep_flag": False},
}

The CrewAI tool schema is automatically generated from EvaluationInput — agents can construct the correct input from natural language instructions.