How to find verified work emails with an API

You get a work email your agent can trust by calling an API that returns the address with a source and a confidence score — not a bare string — then setting a threshold the agent won't act below. That's the short answer. Here's why the distinction matters.

"How can my agent find verified work emails?" is really two questions stacked, and the second is the one that trips agents up. A pattern-guessed address — [email protected] — is a guess, not a verified email. It might bounce, and an agent that sends to a bounce burns a real prospect and dents your sending reputation. So the useful version of the question is narrower: how do you get a work email an autonomous system can act on without a human eyeballing it first?

Find and verify are different jobs

• Finding is: given a name and a company (or a profile URL), produce the most likely address. Every provider can guess a pattern. The better ones cross-reference real sources instead of just assembling first.last@.
• Verifying is: checking that the address actually resolves — MX records, an SMTP handshake, provider signals — so you know it's deliverable, not merely plausible.

An API that does both and tells you how sure it is is the difference between "here's an email" and "here's an email you can safely send to."

What "verified" honestly means — and what it doesn't

No finder is right every time, and anyone who claims otherwise is selling. People change jobs between the crawl and your call. Plenty of domains run catch-all servers that accept any address, so a clean SMTP check proves nothing. Some mailboxes can't be checked at all without actually sending. So the honest output isn't a boolean "verified ✓" — it's a confidence score and the source behind it, so the caller can set its own bar.

That's exactly what makes it usable by an agent. You can tell it: act only above 0.8; below that, flag for a human. A bare string gives an agent no way to make that call — it has to trust all of it or none of it.

How to wire it for an agent

abm.dev is the account-based marketing API for AI agents, and it returns every field — a work email included — with its source and a confidence score, cited or not returned. So an agent gets the address, where it came from, and how sure the system is, in one call. Set a confidence threshold in your prompt or your code and let the agent decide when to use the email and when to hold. One key, over MCP or REST — no stitching a finder and a verifier together yourself.

The test

Ask an agent to find someone's work email and look at what comes back. If it's a bare address, you're getting a guess dressed as an answer. If it comes back with a source and a confidence score, you're getting something an autonomous system can actually reason about — and that's the only kind of "verified" worth the word.

Try it

The playground is free — one key gets you the REST API or the MCP server at https://mcp.abm.dev/mcp. Point an agent at it and check the shape of the answer, not just the address.