If you’ve got a spreadsheet of emails collected from a webinar, a lead magnet, or a scraped list, this tutorial will show you how to hand the whole cleanup job to an AI agent using RocketVerifier’s MCP server. No API code, no scripts — just a conversation.

What you’ll need

  • A RocketVerifier account — sign up free for 100 credits
  • An API key from the dashboard
  • Claude Desktop or Cursor (or any other MCP-compatible client)

Step 1: Connect the MCP server

Open your client’s MCP configuration and add RocketVerifier:

{
  "mcpServers": {
    "rocketverifier": {
      "command": "npx",
      "args": ["-y", "@rocketverifier/mcp"],
      "env": {
        "ROCKETVERIFIER_API_KEY": "rv_live_your_api_key"
      }
    }
  }
}

Restart the client. You should see rocketverifier appear in its list of connected tools, exposing verify_email, create_bulk_job, get_job_results, check_credits, and friends.

Step 2: Sanity-check the connection

Before touching real data, ask your agent:

“Check my RocketVerifier credit balance.”

If it responds with an actual number, the connection is live. This calls the check_credits tool under the hood — no credits consumed.

Step 3: Hand over the list

Paste your list directly into the chat, or point the agent at a file:

“Here’s a list of 40 emails from our last webinar signup. Verify each one, tell me which are risky or invalid, and give me a clean CSV of just the valid ones.”

The agent will call verify_email (or batch through create_bulk_job for larger lists) once per address, read back the status, score, and reason for each, and assemble the answer in the format you asked for — a table, a CSV block, a short summary, whatever you prefer.

Step 4: Go bulk for large lists

For lists beyond a few hundred addresses, ask the agent to use the bulk job tools instead of one-by-one calls:

“Create a bulk verification job for this list of 5,000 emails, then poll until it’s done and summarize the results by risk category.”

Under the hood this chains create_bulk_job → get_job_status (polled) → get_job_results, which is dramatically more efficient than thousands of individual verify_email calls and won’t burn through your rate limits.

Step 5: Ask follow-up questions naturally

Because the whole thing lives inside a conversation, you can keep iterating:

“Of the invalid ones, how many are disposable vs. just bad syntax?”

“Re-check just the ones marked ‘risky’ with a longer timeout.”

“Export everything with a confidence score above 80 as JSON.”

The agent has full context from the previous tool calls, so these follow-ups just work.

Try it risk-free

Every tool also works against sandbox addresses (user@sandbox.rocketverifier.com) with a test API key (rv_test_...), so you can run through this entire workflow without spending real credits first.

Where to go next