CrewAI

Installation

pip install crewai requests

Basic tool definition

import os
import requests
from crewai.tools import BaseTool
from pydantic import BaseModel, Field

class EmailVerifyInput(BaseModel):
    email: str = Field(description="The email address to verify")

class VerifyEmailTool(BaseTool):
    name: str = "verify_email"
    description: str = (
        "Verify if an email address is real and deliverable. "
        "Returns valid status, confidence score, and whether "
        "the address is disposable or a role address."
    )
    args_schema: type[BaseModel] = EmailVerifyInput

    def _run(self, email: str) -> dict:
        response = requests.post(
            "https://api.truval.dev/v1/email/verify",
            headers={
                "Authorization": f"Bearer {os.environ['TRUVAL_API_KEY']}",
                "Content-Type": "application/json",
            },
            json={"email": email},
        )
        return response.json()

Building an autonomous verifier lead agent

An agent shouldn’t just run the tool; it should make decisions based on the result. Here’s how to create an agent that uses the signals to filter bad leads.

from crewai import Agent, Task, Crew

verify_tool = VerifyEmailTool()

lead_enricher = Agent(
    role="Lead Quality Validator",
    goal="Filter out bad emails before reaching out to leads. Never approve disposable or invalid emails.",
    backstory="You are a strict QA agent. You use verify_email. Reject if disposable or status is invalid/undeliverable. If smtp_blocked with confidence ~0.75, approve domain-level check. If status is unknown with mx_found and not smtp_blocked (often 0.50), ask for confirmation — do not auto-reject. Reject only if confidence is very low after those rules (e.g. undeliverable).",
    tools=[verify_tool],
    verbose=True,
)

task = Task(
    description=(
        "You have a new lead with email: hello@mailinator.com. "
        "Verify this email. If it's valid with high confidence and not disposable, approve it. "
        "Otherwise, reject it and explain why."
    ),
    expected_output="A final decision: 'APPROVED' or 'REJECTED: [Reason]'",
    agent=lead_enricher,
)

crew = Crew(agents=[lead_enricher], tasks=[task])
result = crew.kickoff()

# The agent autonomously executes the tool, gets `disposable: true`,
# and returns "REJECTED: Email is a disposable address."

Environment variables

export TRUVAL_API_KEY=sk_live_...

Get your API key at dash.truval.dev.

Public Repositories

• SDK: truval-dev/truval-sdk
• MCP Server: truval-dev/truval-mcp-server
• Agent Skills: truval-dev/truval-skills