GTM-MCP — B2B Cold Outreach for AI Agents

An open-source MCP server that turns Claude Code into a full B2B cold outreach pipeline. Tell it what you sell and who you're targeting — it finds companies, verifies they're a fit, extracts contacts, writes email sequences, and creates ready-to-send campaigns.

What is Cold Outreach?

You have a product. There are companies out there that need it but don't know you exist. Cold outreach is how you reach them — you send a message to someone who's never heard of you, explaining how you can help.

Outreach channels include email, LinkedIn, WhatsApp, Telegram, and cold calls. This MCP focuses on email (LinkedIn via GetSales is also supported).

The Math

Cold outreach works on volume. Even with perfect targeting and great emails, expect roughly 1 positive reply per 100 people you contact. That's normal — most people ignore cold emails. But if you contact 1,000 relevant people, that's ~10 interested leads. At 5,000 that's ~50.

So you need thousands of verified, relevant contacts. You can't do this manually — you need a pipeline.

The Pipeline: How Email Outreach Works

Here's the full journey from "I have a product" to "replies in my inbox":

1. Find companies (data provider)

You need a database of millions of companies to search through. Apollo is one such provider — you search by industry, location, company size, keywords, and it returns thousands of results.

The problem: Apollo's filters are broad. Searching "fintech companies in US" returns thousands of results, but maybe only 30-40% actually need what you sell. Apollo gives you the raw material, not the finished list.

2. Verify they're actually your target (scraping + AI)

This is where most outreach fails — people skip this step and blast emails to everyone Apollo returns. The result: low reply rates and your domain gets flagged as spam.

Instead, this pipeline scrapes each company's website and has your AI agent read it: "Based on what this company does, would they actually need our product?" This filters out 60-70% of Apollo results, leaving only real targets.

For website scraping, we use Apify — specifically its residential proxies, which make requests look like regular browser traffic so websites don't block you.

3. Find the right people at each company

At each verified target company, you need the right person's email — typically a decision-maker (CEO, VP Sales, Head of Partnerships, etc.). Apollo provides this too: first a FREE people search to find names, then 1 credit per person for a verified email address.

4. Write email sequences

You don't send one email — you send a sequence: an initial email, then 2-3 follow-ups spaced days apart. Each email should be short, plain text (no HTML, no images, no tracking links — these trigger spam filters), and focused on the value you provide.

The best emails reference real case studies: "We helped [similar company] achieve [specific result]" converts much better than generic pitches.

5. Send via an email sequencer

You can't send cold emails from Gmail — Google will shut you down. Instead, you use a sequencer like SmartLead that handles:

• Sending from multiple domains/inboxes to stay under spam thresholds
• Scheduling follow-ups automatically
• Tracking who replied, who opened, who bounced
• Warming up new email accounts (see below)

6. Monitor replies and follow up

When people reply, you triage: interested (hand to sales), not interested (stop emailing), out of office (wait and retry). This MCP also automates reply classification.

Email Infrastructure You Need (~2 Weeks Setup, Rotate Every 4-6 Weeks)

Before sending cold emails, you need sending infrastructure. Initial setup takes ~2 weeks. Domains degrade over time as recipients mark emails as spam, so you'll need to rotate — buy new domains, warm them up, and retire old ones. Typica

... View full README on GitHub