Most agents can already write the cold email. Almost none of them can answer the only question that matters before you hit send: is this address real?
TAPAC (tapacapi.com) is a B2B contact engine you wire into an agent over MCP. As of this week it installs in one line:
npx -y @tapacapi/mcp
What it does
Two tools, no ceremony:
| Tool | Purpose |
|---|---|
tapac_find_contacts |
Find and verify contacts. Params: industry, job_titles, company_size, location, source (website \ |
tapac_status |
Server version, API-key state, live API reachability. |
Every contact comes back with name, title, company, email, source, and SMTP verification status. The server never invents contacts — it returns exactly what the API returns.
Wire it into your agent
{
"mcpServers": {
"tapac": {
"command": "npx",
"args": ["-y", "@tapacapi/mcp"],
"env": { "TAPAC_API_KEY": "your_key_here" }
}
}
}
Prefer Python? The same server ships as a uvx install with stdio and streamable-HTTP transports:
uvx --from git+https://github.com/axelfreeman/tapac-mcp tapac-mcp
Grab a free key — 100 searches included — at tapacapi.com/get-key. Without a key the server still starts and returns onboarding instructions instead of contacts, so you can wire it up before you have a key (that's how I tested the npm build: tapac_status reported the key as unset and pointed at the key page).
Why verification beats databases
The numbers that pushed me here, with sources:
- 23% of contacts change jobs annually (ZoomInfo 2025) — a list that was accurate in January is leaking by June.
- 40% of emails are dead within two years (NeverBounce).
- Fresh scraped data outperforms stale databases by 42% (Harvard Business Review 2024).
- Static databases bounce at 10–35%; TAPAC verifies each address over SMTP before it hands it over, so expected bounce is 2–5%.
Bounce rate is not a vanity metric: every hard bounce is a reputation hit on the sending domain your whole outbound motion depends on. A 30% bounce list doesn't just waste sends, it teaches every mailbox provider that your domain is noise.
TAPAC is pay-per-use at $0.10–0.50 per contact — you pay for contacts you actually get, instead of a seat licence for a database that resells the same records to your competitors.
Three agent workflows worth building
-
"Find 20 VP Sales at US SaaS companies with 50–500 employees"
tapac_find_contacts(industry="SaaS", job_titles=["VP Sales"], company_size="50-500 employees", location="US", limit=20) -
Community sourcing —
tapac_find_contacts(source="telegram")orsource="discord"when the buyers live in chat, not on a landing page. Community-sourced leads convert 30% better than cold outreach (McKinsey 2025). -
Pre-send gate for any outbound agent — call
tapac_find_contacts(or the SDK's verify path) and drop every address that fails SMTP verification before the email tool is ever invoked. Personalised email gets 3.2x higher open rates (HubSpot 2025), and personalisation is wasted on a dead address.
Links
- Product and docs: tapacapi.com
- MCP server, SKILL.md and agent manifest: github.com/axelfreeman/tapac-mcp
- npm:
@tapacapi/mcp· PyPI SDK:tapac-sdk· npm SDK:@tapacapi/sdk
Facts above are cited from the sources named inline and from TAPAC's own published agent instructions. Nothing here is a benchmark I invented — if a number has no source, it isn't in this post.
Top comments (0)