Smartlead MCP Server: Connect Smartlead to Claude (2026 Guide)

The Smartlead MCP server lets Claude read and run your Smartlead campaigns from chat. You can ask for a weekly reply-rate report, flag a mailbox with high bounces, or upload a lead list without opening the dashboard.

Setup takes about five minutes, but the docs are scattered. Smartlead's help center, its own blog, and community servers each describe a different install path.

This guide gives you the working setup, the actions you get, copy-paste prompts, the real rate limits, and the guardrails to add before Claude touches a live campaign. It ends with how SyncGTM enriches and verifies leads before they ever reach Smartlead.

What Is the Smartlead MCP Server?

The Smartlead MCP server is a bridge between Claude and the Smartlead API, built on the Model Context Protocol. Claude sees Smartlead actions as tools and calls them when your prompt needs live data.

Smartlead's own blog says the server exposes 116+ tools across six areas. Those are campaign management, lead lifecycle, email accounts, deliverability diagnostics, analytics, and webhook automation.

Every call is a real API request against your account. Nothing is cached, so results are current, and every call counts against your rate limit.

If you are still deciding whether Smartlead is the right sending tool, read our Smartlead review first.

Official vs Community Smartlead MCP Servers

Three routes show up when you search for a Smartlead MCP. The official server is the one to use unless you need a feature it lacks.

One trap: Smartlead's newer setup blog references an @smartlead/mcp-serverpackage. When we checked the public npm registry on October 18, 2026, that name returned no package, so this guide uses the help center's mcp-remote route.

How to Connect Smartlead to Claude

The official Smartlead MCP connects in five steps. You need a paid Smartlead plan with API access, Claude Desktop, and Node.js LTS.

Step 1: Get your Smartlead API key

In Smartlead, open Settings → API Keys and generate or copy a key. This key grants full API access, so never paste it into a shared doc or chat.

Step 2: Open the Claude Desktop config

In Claude Desktop, go to Settings → Developer → Edit Config. That opens claude_desktop_config.json.

On Mac it lives at ~/Library/Application Support/Claude/. On Windows it lives at %APPDATA%\Claude\.

Step 3: Add the Smartlead server block

Add this entry under mcpServers, per Smartlead's MCP help article. Keep any servers already in the file.

{
  "mcpServers": {
    "smartlead": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.smartlead.ai/sse?user_api_key=YOUR_API_KEY"
      ]
    }
  }
}

The mcp-remote package turns the remote SSE endpoint into a local server Claude Desktop can launch. Check the key in a plain-text editor first, because a trailing space breaks authentication.

Step 4: Restart Claude Desktop

Fully quit Claude Desktop, not just the window, then reopen it. The first launch can take 10 to 20 seconds while npx downloads mcp-remote.

Step 5: Enable the tools and test

Open a new chat, click the tools icon, and make sure Smartlead is toggled on. Start with a read-only prompt so nothing changes on a live campaign.

A real call shows a tool-use block in the reply. A table with no tool block means Claude answered from memory, so ask it to use the Smartlead tools.

Optional: Claude Code

Smartlead only documents Claude Desktop. Claude Code still accepts SSE servers through its MCP command, so this pattern should work, though Smartlead does not support it officially.

claude mcp add --transport sse smartlead \
  "https://mcp.smartlead.ai/sse?user_api_key=YOUR_API_KEY"

Smartlead MCP Campaign and Lead Actions

The Smartlead MCP groups its tools into six areas. Read actions are safe to run freely, while write actions change live data.

Smartlead names tools like get_campaign_status, fetch_lead_data, and check_deliverability. Ask Claude “list every Smartlead tool you can call” to see the exact set in your account.

Example Prompts for Reporting and Lead Uploads

Good Smartlead MCP prompts name the time window, the metric, and the output format. Here are prompts grouped by job.

Reporting

• “Summarize my top 5 campaigns by reply rate for the last 30 days as a table with sent, replies, and interested replies.”
• “Which sequence step gets the most replies in campaign [name]? Show reply rate per step.”
• “Write a Monday report: sends, replies, interested leads, and the biggest week-over-week drop.”

Deliverability checks

• “Show any mailbox with a bounce rate above 5% in the last 7 days.”
• “List disconnected email accounts and any account where warmup has stopped.”

Pair these with a warmup plan, since MCP only reports problems. Our email domain warmup guide covers the fix side.

Lead uploads

• “Add these 50 leads to campaign [name]. Map first name, company, and title to custom fields. Show me the payload before you upload.”
• “Find leads who replied in the last 7 days but are still active in a sequence, then list them. Do not change anything yet.”

Note the “show me first” wording. It turns a write action into a two-step review, which matters more than any other prompt habit.

Smartlead MCP Rate Limits

The Smartlead MCP does not have its own quota; it inherits your API limits. Smartlead's rate-limit guide sets them per API key across all endpoints.

Three endpoints add a stricter cap of 10 requests per 60 seconds each: campaign leads statistics, all-leads activities, and domain-wise mailbox health. A prompt like “show lead stats for all 15 campaigns” will hit that wall after 10 calls.

A 429 error on the plan limit returns a Retry-After value. The endpoint-specific 429 does not, so wait a full 60 seconds.

For lead uploads, the current guide allows 400 leads per request, while an older reference page says 350. Ask Claude to batch at 350 and you stay inside both.

Practical rule: one campaign report is a handful of calls, but “audit every campaign” can be 100+. Scope prompts to a few campaigns at a time.

Safety Guardrails for Write Actions

The Smartlead MCP can pause campaigns, change lead status, edit sequences, and upload leads. Smartlead's docs give no confirmation guidance, so set these guardrails yourself.

1. Keep approval prompts on.Claude Desktop asks before each tool call. Choose “Allow once” for write tools and save “Allow always” for read tools only.
2. Deny destructive tools in Claude Code. Add delete and status tools to the permissions.deny list in settings, using the mcp__smartlead__<tool_name> pattern.
3. Test on a sandbox campaign. Create a paused campaign with test leads, and run every new write prompt there first.
4. Ask for a dry run.End write prompts with “show me the changes before you make them.”
5. Protect the key. The key sits in the URL query string, so it can land in logs. Never share your config file, and regenerate the key if it leaks.
6. Pause before editing. Pause a live campaign before Claude changes its sequence, then resume it after review.

Troubleshooting Common Errors

For other local-server issues, Anthropic's Claude Desktop MCP guide explains where logs live.

Limitations of the Smartlead MCP

The Smartlead MCP is strong at operating campaigns. It is not built to decide who goes into them.

• No lead sourcing. It cannot search for new companies or people by ICP.
• No email finding or verification. It uploads whatever addresses you give it, valid or not.
• Desktop only, SSE only. SSE is deprecated in newer MCP spec versions, and Claude Web support is still listed as planned.
• Tight analytics caps. The 10-per-minute endpoints make large reports slow.
• Broad key scope. One API key grants full account access, with no read-only option documented.

The second gap costs the most. Bad emails raise bounce rates, and high bounces damage the sender reputation Smartlead's warmup works to build.

A Better Way: Enrich With SyncGTM Before Smartlead

The better workflow runs two MCPs in the same Claude chat. SyncGTM's MCP builds and verifies the list, then the Smartlead MCP sends to it.

How the workflow runs

1. Find people. Claude calls find_people with your ICP filters, such as title, company size, and country. Cost: 0.2 credits per person returned.
2. Find work emails. find_work_email tries 50+ providers in a waterfall until one matches. Cost: 1 credit, and misses are not billed.
3. Verify every email. verify_email checks deliverability across multiple verifiers. Cost: 0.1 credits.
4. Add a reason to reach out. Optional signals like job changes (2 credits) or company firmographics (2 credits) feed personalization fields.
5. Upload to Smartlead. Claude passes only verified leads to the Smartlead MCP in batches of 350, or the SyncGTM Smartlead integration pushes them into a campaign directly.

What 100 verified leads cost

See SyncGTM pricing for what a credit costs on each plan. Run check_credits in Claude before a big list.

Connect SyncGTM's MCP

Unlike Smartlead's setup, SyncGTM uses a remote HTTP server with OAuth, so there is no API key in a config file. In claude.ai, open Settings → Connectors → Add custom connector and paste https://api.syncgtm.com/mcp.

# Claude Code
claude mcp add --transport http syncgtm https://api.syncgtm.com/mcp

Our lead generation with Claude guide walks through the full setup. For the native, no-chat route, see the SyncGTM + Smartlead integration.

Conclusion

The Smartlead MCP server is the fastest way to run Smartlead reporting and campaign operations from Claude. Use the official mcp-remote setup, scope prompts to stay under 60 calls a minute, and keep approval prompts on for every write tool.

It does not choose or clean your leads. Put SyncGTM's MCP in front of it, and every lead that reaches a sequence has a found, verified email.

Comparing sending tools too? Our Smartlead alternatives roundup and top GTM MCPs for Claude Code cover the rest of the stack.