Email Finder MCP: Find Verified B2B Emails in Claude (2026)

An email finder MCPlets Claude look up a prospect's verified work or personal email without you ever opening a prospecting tool. You describe the person in a sentence. Claude calls the lookup tool, gets a sourced address back, and keeps going with whatever you were actually doing.

The catch is that most email finder MCP servers wrap a single data source. When that one source misses, you get nothing — and you often only find out after the bounce.

This guide wires up the SyncGTM Lead Gen and Enrichment MCP, which runs a waterfall across 50+ providers instead. Setup takes one command, and by the end you will be enriching a full CSV from a single Claude prompt.

What Is an Email Finder MCP?

An email finder MCP is a Model Context Protocol server that exposes email-lookup tools to an AI client. MCP is the open standard Anthropic published so models can call external services through a typed, permissioned interface.

In practice, that means the email finder stops being a destination. It becomes a capability Claude reaches for mid-task, the same way it reaches for a file read.

You type "find me the work email for the VP of Engineering at Acme." Claude picks the right tool, fills the arguments from context, executes, and hands back a structured result you can drop straight into a sequence.

Why an Email Finder MCP Beats a Single-Source Lookup

Every email database has holes. One vendor is strong on US mid-market SaaS, another on European manufacturing, another on recent job-changers.

A single-source lookup inherits exactly one of those coverage profiles. Whatever it misses, you miss.

Waterfall enrichment queries providers in order and stops at the first hit. The reps who used to keep three tabs open — one for Hunter, one for Apollo, one for a scraper — were running a waterfall by hand. The MCP just does it in a single tool call.

The MCP layer adds a second thing manual tab-hopping never had: composition. Claude can chain find_people into find_work_email into verify_email without you defining a pipeline.

If you are weighing standalone tools instead, our ranked B2B email finder comparison and Hunter.io alternatives cover the non-MCP options.

Prerequisites

The setup is short. You need four things before you start.

• An MCP-capable Claude client — Claude Code, Claude Desktop, Cursor, or Windsurf. This guide uses Claude Code because the install is a single terminal command.
• A SyncGTM account with credits on it. The OAuth flow signs you into this account, so create it first.
• Prospect inputs — at minimum a full name plus a company domain or LinkedIn URL. Name alone is not enough for a reliable match.
• Somewhere for the output to land — a CSV, a CRM, or a sequencer. Decide this now so Claude writes results where you need them.

How to Set Up an Email Finder MCP in Claude

Six steps. The first two are the install; the rest is the workflow you will actually run every day.

Step 1: Add the MCP Server

SyncGTM ships as a remote HTTP server, so there is nothing to clone, build, or keep running locally.

# Add to the current project
claude mcp add --transport http syncgtm https://api.syncgtm.com/mcp

# Or make it available in every project
claude mcp add --scope user --transport http syncgtm https://api.syncgtm.com/mcp

Use --scope user if prospecting is something you do across repos. Skip it if you want the server scoped to one working directory.

Setup instructions for Claude Desktop, Cursor, and Windsurf live in the SyncGTM MCP documentation.

Step 2: Authenticate With OAuth

Restart Claude Code, then run the MCP command and pick the server.

# Inside Claude Code
/mcp

# Select "syncgtm" → "Authenticate"
# Your browser opens the SyncGTM sign-in page
# Approve, and the session is stored automatically

There is no API key to paste and no token to rotate. That matters more than it sounds — a key in a config file is a key that eventually ends up in a commit.

Confirm the connection worked by asking Claude to run check_credits. If it returns your balance, every enrichment tool is live.

Step 3: Find Your First Work Email

No syntax to learn. Describe the person and let Claude select the tool.

You: Find the work email for Priya Raman, VP of Marketing at
     figma.com. Use the SyncGTM MCP.

Claude: [calls find_work_email]
        → priya.raman@figma.com

find_work_email takes the person's name plus a company domain or LinkedIn URL. It runs the waterfall and returns the first sourced address it finds.

Give it the domain, not the company name, whenever you have it. "Acme" matches a dozen companies; acme.io matches one.

Step 4: Add the Personal-Email Fallback

Some contacts have no discoverable company mailbox. Founders between roles, candidates, and people at very small firms are the usual cases.

That is what find_personal_email is for. Tell Claude the fallback rule once and it applies it to everything after.

You: For every contact, try find_work_email first. If it returns
     nothing, fall back to find_personal_email. Flag which one
     was used in a "source" column.

Keep the two apart in your data. A personal address that gets loaded into a work-email sequence reads as a privacy violation to the recipient, and it is a fast route to a spam complaint.

Step 5: Verify Before You Send

Found is not the same as deliverable. People leave, mailboxes get deprovisioned, and stale records outlive the person they belonged to.

verify_email is a separate check, and it is the cheapest insurance in the stack at 0.3 credits per row.

You: Run verify_email on all 40 addresses. Drop anything that
     is not deliverable, and give me the pass rate.

Note the billing difference: verification is charged per row, not per result. You pay for the check whether the address passes or fails.

Step 6: Go From One Contact to a Full List

One lookup is a demo. The workflow that saves real hours is the batch.

You: Read prospects.csv. For each row use the "name" and
     "company_domain" columns to find a work email, verify it,
     then write email + status back as new columns. Start with
     the first 25 rows and report the hit rate.

Always run a sample batch first. A 25-row test tells you whether your input data is good enough before you spend credits on 2,000 rows.

If you do not have a list yet, source one in the same session. find_companies returns accounts matching your ICP filters, find_people returns the contacts inside them, and enrich_organization fills in firmographics. The list-building and the enrichment happen in one conversation.

For the wider agent pattern — research, enrichment, and sequence drafting in one loop — see Claude Code for sales prospecting.

Which Tools Run, and What They Cost

The SyncGTM MCP exposes roughly 50 tools. These are the ones that matter for email work, with their credit cost and billing unit.

The distinction between the two billing units is the one to internalise. Per result means you are charged only when data comes back, so a miss on find_work_email is free. Per row means you pay for the attempt.

Run the numbers on your own volume with the credits estimator, or compare plan tiers on pricing.

Tips and Best Practices

Feed it domains, not company names

Match quality is bounded by input quality. A LinkedIn URL or a company domain removes the ambiguity that a name alone leaves behind.

Always verify before the address hits a sending tool

Bounce rate drives domain reputation, and reputation is expensive to rebuild. At 0.3 credits per row, verification is the cheapest line in the entire workflow.

Sample before you commit

Run 25 rows and check the hit rate. If it comes back low, the problem is almost always the input columns — not the waterfall.

Put the workflow rules in CLAUDE.md

Write the fallback order, the verification rule, and your output schema into a project file. Claude then applies them every session without you restating them.

Reserve mobile lookups for the shortlist

At 12 credits per result, find_mobile_number is the most expensive call in the set. Run it on replies and hand-picked accounts, never on a cold list.

Respect consent rules

GDPR, CAN-SPAM, and CASL apply to an MCP-sourced address exactly as they apply to a manually sourced one. Automation changes the speed, not the obligation.

Troubleshooting

One more thing worth knowing: because the addresses come back as structured tool results rather than scraped text, they can flow straight into a CRM write or a sequencer without a cleanup pass. That is the practical difference between an MCP and a browser extension.

Conclusion

An email finder MCP removes the worst part of prospecting: the context switch. The lookup happens where you are already working, in the same sentence as the request.

What separates a useful one from a novelty is the data underneath. A pattern guess confirmed by SMTP will fail on every catch-all domain you point it at, and a single-provider wrapper dead-ends at that provider's coverage limit.

A waterfall keeps going. That is the whole argument, and it is why find_work_email being charged per result rather than per attempt is more than a billing detail — you only pay when the waterfall actually delivers.

Install it with one command, authenticate in the browser, and run 25 rows to see your own hit rate. If you want the wider picture first, read Claude data enrichment for the full 50-provider workflow, or the best waterfall email finders for how the underlying approach compares outside MCP.