GTM-engineering walkthrough · Smartlead MCP · 2026

Smartlead + Claude via LeadMagic's MCP — granular cold-email execution in your AI client

Smartlead's MCP layer is built by LeadMagic (Smartlead's official partner) at github.com/LeadMagic/smartlead-mcp-server. It exposes 113-116 granular tools across the full Smartlead surface and runs as a stdio Node process via npx. For technical operators who want maximum control over the deliverability-heavy cold-email workflow from an AI client, this is the structural shape that makes Smartlead the LLM-native power-user pick in 2026.

This page is the operator-grade walkthrough: ~10-minute setup, five concrete workflows that leverage the granular tool surface, the disambiguation gotcha (113 tools means occasional wrong-tool routing), and the multi-MCP source-to-send pattern (Apollo + Smartlead + HubSpot in one Claude session).

Transport
stdio
Local Node process via npx
Auth
SMARTLEAD_API_KEY
Env var in MCP client config
Setup time
~10 min
Node + npx + env var + restart
Tools
113-116
Granular, full operations coverage

TL;DR

Want to try Smartlead?

Wire Smartlead into Claude with LeadMagic's 113-tool MCP and ship power-user cold email

Granular operations coverage, official-partner maintained, free open-source. Pair with Apollo + HubSpot MCP for the full source-to-send loop.

Start with Smartlead →Affiliate link — StackSwap earns a commission if you sign up for Smartlead. We only partner with tools we'd recommend anyway.

Five workflows that leverage the granular surface

1. Cross-campaign analytics synthesis

"Compare reply rate across all Q2 campaigns by sender pool. Surface sequence steps with greater than 15% reply-rate drift from the 30-day baseline. Group senders by warmup vintage and SMTP provider." Claude pulls per-campaign and per-sender stats via the granular MCP tools, aggregates, and returns a structured analysis. Replaces the multi-tab Smartlead UI session that would take an hour.

2. Sender-pool health management

"List all warmup-state senders with bounce rate above 3%. Pause them. Identify the common pattern — same SMTP, same DKIM config, same warmup vintage." The granular tool surface lets the LLM fetch per-sender state, execute the pause on the right scope, and analyze the pattern in one conversation.

3. Lead-routing on positive reply

"Find leads with positive replies across all active campaigns from the past 7 days. Route them to the AE follow-up sequence with personalization grounded in the original reply." Claude fetches replies across campaigns, classifies positive intent, executes the lead-move + personalization-population in one pass.

4. Sequence variant testing

"Create an A/B variant of step 2 in the enterprise campaign with a shorter, more question-led subject line. Route 50% of new leads to the variant. Schedule a performance review prompt for me in 14 days." The granular tools let the LLM orchestrate sequence configuration, traffic split, and scheduled follow-up in one conversation.

5. Deliverability incident triage

When a sender pool drops in reply rate, ask Claude to fan out across affected senders, identify the pattern (SPF/DKIM misalignment, content trigger, infrastructure issue), and recommend the fix. The granular tool surface is essential here — Claude needs to query per-sender SMTP config, per-sender bounce reasons, per-sender content patterns. A coarse-grained MCP wouldn't expose enough surface to do this triage.

Setup — ~10 minutes

  1. Install Node.js if you don't have it (18+ recommended).
  2. Generate a Smartlead API key. Smartlead Settings → API. Scope to a dedicated AI user if possible.
  3. Edit Claude Desktop config. Add an MCP server entry under mcpServers that invokes smartlead-mcp-server via npx with SMARTLEAD_API_KEY env var. Repo README has the exact snippet.
  4. Restart Claude Desktop. The 113+ tools appear in your next session.
  5. Verify connectivity. Ask Claude "list my active Smartlead campaigns".

The granular-tool gotcha

With 113 tools the LLM occasionally picks the wrong granular tool. "Pause this campaign" might route to several different pause operations. Mitigations:

Want to try Smartlead?

Smartlead + LeadMagic's MCP is the granular control plane for AI-driven cold email in 2026

113-tool granularity, full operations coverage, official-partner maintained. The right shape for technical operators.

Start with Smartlead →Affiliate link — StackSwap earns a commission if you sign up for Smartlead. We only partner with tools we'd recommend anyway.

FAQ

Smartlead MCP runs as a stdio Node process via npx. (1) Install Node.js (18+) if you don't have it. (2) Generate a Smartlead API key in Settings → API. (3) Edit your Claude Desktop config (claude_desktop_config.json) to add an MCP server entry that runs the smartlead-mcp-server package via npx with SMARTLEAD_API_KEY as an env var. (4) Restart Claude Desktop. (5) Verify connectivity — 'list my active Smartlead campaigns'. The repo at github.com/LeadMagic/smartlead-mcp-server has the exact config snippet in the README.

You need a Smartlead account with API access (any paid tier qualifies). Smartlead plans start around $39/mo. The MCP server itself is free — LeadMagic maintains the open-source project. You're paying for Smartlead, not for the MCP layer.

Five validated patterns: (1) cross-campaign analytics — 'compare reply rate across all Q2 campaigns by sender pool, surface steps with >15% reply-rate drift'; (2) sender-pool health management — 'list all warmup senders with bounce >3%, pause them, identify the SMTP pattern'; (3) lead-routing on positive reply — 'find leads with positive replies across all campaigns, route to AE follow-up sequence with personalization'; (4) sequence variant testing — 'create A/B variant of step 2 with shorter subject, route 50% of new leads'; (5) deliverability incident triage — when a sender pool drops, fan out across affected senders, identify the pattern (SPF/DKIM, content trigger, infrastructure), recommend fix. The 113-tool granularity is what makes these workflows precise.

Yes — this is the source-to-send-to-CRM loop MCP enables. Configure Claude Desktop with Apollo MCP (sourcing), Smartlead MCP (execution), and HubSpot MCP (CRM updates). The LLM orchestrates: pull prospects from Apollo, push to a Smartlead campaign with per-row personalization, route positive replies back to HubSpot. Each MCP is a separate entry in claude_desktop_config.json — they don't conflict.

Order-of-magnitude friction difference. Zapier path: AI client invokes webhook → Zapier receives → transforms → calls Smartlead API → response back. Each hop adds latency, Zapier seat is $20-200+/mo. Smartlead MCP path: AI client invokes Smartlead MCP directly → response. One hop. Free. LeadMagic-maintained (so the integration tracks Smartlead's API closely).

Limited. If your team runs Smartlead through the UI and your daily orchestration doesn't run through Claude / Cursor / ChatGPT, the MCP layer is overhead. Smartlead's product earns its $39+/mo on the deliverability + sequencing engine; MCP is incremental leverage for AI-forward operators specifically.

Real but manageable. With 113 tools the LLM occasionally picks the wrong granular tool for a vague request. Mitigations: (1) be explicit in prompts ('use pause_campaign on ID 123'); (2) system-prompt the disambiguation rule ('when user says pause, clarify scope before acting'); (3) trust the granularity for power workflows — for complex multi-step operations, the granular surface is the asset.

It's an official-partner project, not a first-party Smartlead server. LeadMagic is Smartlead's official partner, the MCP is linked from Smartlead's documentation at helpcenter.smartlead.ai, and Smartlead's team is aware of and supportive of the project. There's no formal SLA on the MCP server itself, but the project has been actively maintained throughout 2025-2026.

Related reading

Canonical URL: https://stackswap.ai/smartlead-mcp-claude-integration. Disclosure: StackSwap is a Smartlead affiliate.