How to Set Up lemlist MCP in 2 Minutes (Claude, Cursor & More)

Last week I asked Claude to pull reply rates on three campaigns, find the one that was leaking, and draft two new follow-ups. I never opened lemlist. The whole thing took four minutes, and two of those were me re-reading the drafts because I didn’t trust them yet. That’s what happens when your AI client can talk to your outbound stack directly, through the Model Context Protocol that Anthropic announced in November 2024. No exports. No copy-paste. No tab-switching.

We built lemlist’s MCP server so you never have to paste a lead list into an AI chat again. In this guide, I’ll show you exactly how to connect it, OAuth or API key, in under 2 minutes. Then I’ll show you the tool categories you unlock, how to scope the connector down when you want a focused agent, and the exact prompt we use to turn Claude Code into a dedicated lemlist operator.

There are 2 authentication methods

lemlist’s MCP server supports two ways to authenticate.

OAuth (recommended). This is the simpler path. Your AI client handles the entire authentication process automatically: it opens a browser window, you log in and grant access to your lemlist team, and that’s it. No key management, no copy-pasting credentials. Access tokens are valid for 1 hour and refresh automatically for up to 30 days. (Yes, the token refresh happens quietly in the background, no more re-pasting keys every hour.)

API key. If your client doesn’t support OAuth flows, or you prefer to manage credentials manually, you can authenticate with a lemlist API key. The key is generated once inside lemlist and pasted into your client’s config file. Best for Cursor, CI environments, or custom integrations.

We recommend OAuth for most teams, because the way you authenticate is the thing you’ll stop thinking about five minutes from now, and OAuth is the option that disappears fastest. Outbound is turning prompt-driven. The less friction between a question in your head and an answer from your campaign data, the more of your day goes to judgment instead of clicking.

⚠ Important note: the MCP server uses the X-API-Key header for authentication, which is different from the REST API’s Basic Auth method.

Setup with OAuth

Claude Desktop

Open the sidebar, click the + icon next to Connectors, then choose “Add custom connector”.

Fill in:

• •

  Name: lemlist

• •

  URL: https://app.lemlist.com/mcp

Click “Add”. The first time you use a lemlist tool, a browser tab will open asking you to select your team and grant access.

After that, Claude handles token management in the background.

You can also add the config manually. Open claude_desktop_config.json and add:

The mcp-remote package manages the full OAuth flow: client discovery, browser-based consent, PKCE token exchange, and automatic refresh. If you want to know exactly what each of those steps is doing, the MCP specification published by modelcontextprotocol.io is the authoritative reference, defined against the TypeScript schema in schema.ts.

Claude Code

Run this single command in your terminal:

No --header flag needed. On first use, your browser will open a consent page. Once you authorize, you’re connected.

💡 You can set up a custom lemlist agent command in Claude Code. Save the agent prompt into .claude/commands/lemlist-agent.md, then call it with /lemlist-agent to get a pre-configured assistant for campaign management.

Setup with an API key

Step 1 — Generate your lemlist API key

1. 1.

   Log into lemlist and go to app.lemlist.com

2. 2.

   Navigate to Settings > Team > Integrations

3. 3.

   Click “Generate” to create a new API key

4. 4.

   Copy the key and store it somewhere safe. You won’t be able to see it again after closing the page.

Step 2 — Add your key to your client

Cursor

Go to Settings > Tools & MCP > New MCP Server, then add:

Claude Desktop (npm)

Go to Settings/Developer/Edit Config in your app.

Add a “lemlist” block to file claude_desktop_config.json, like in this example:

For Mac users

For Windows users

Restart Claude, and go to the Developer section.

Claude Code

Be careful: always keep your API key out of version control. If you’re working in a shared repo, use environment variables rather than hardcoding the key in your config file.

What you can do once connected

Ask your AI assistant directly: “What lemlist operations can you perform?” It will return the most current list. The available tools evolve daily, so treat anything written here as a snapshot.

Typical categories include:

• •

  Campaign management. Create, edit, pause, or delete campaigns directly from chat.

• •

  Lead management. Update, pause, or remove leads inside a running campaign without touching the UI.

• •

  Lead sourcing. Search lemleads by ICP criteria (role, industry, company size, location).

• •

  Email enrichment. Find and verify emails, and pull in the enrichment options you’d normally trigger by hand.

• •

  Team & stats. Pull open rates, reply rates, and conversion funnels for any time window, across any team you’ve authorized.

• •

  Webhooks. Create and inspect webhook subscriptions so downstream systems stay in sync.

• •

  Message templates & snippets. Draft and reuse multi-step sequence copy with variables and follow-up timing.

The lemlist agent for Claude Code

Once the MCP is connected, you can go further and set up a dedicated lemlist agent command inside Claude Code. This gives you a pre-configured expert assistant that knows how to audit campaigns, source leads, and write sequences without you having to re-explain context every time. If you want that agent to focus on one job, scope its connector with buckets (more on that just below).

How to set it up

Create a file at .claude/commands/lemlist-agent.md inside your project folder. Paste the following prompt into it:

Then open Claude Code and type /lemlist-agent to activate it.

💡 Think of it as a persistent context layer. Every time you call /lemlist-agent, Claude starts the conversation already knowing how to work with lemlist campaigns end to end.

Advanced: scope your connector

By default, your client loads every lemlist tool. That’s fine for exploration. It’s noisy once you know what you want, and it’s a hard sell to a security team that would rather your prospecting agent never see campaign-editing tools at all.

Add a ?bucket= parameter to the endpoint to load only the tool groups you need:

Available buckets: core, builder, prospecting, crm, engagement, deliverability, insights. Comma-separate as many as you want.

Two things this unlocks. Smaller tool surfaces mean fewer wrong-tool calls from the model, which matters more than people expect on long sessions. And you can run several connector entries side by side, one scoped to prospecting for your sourcing agent, another scoped to insights for a read-only reporting agent that can’t touch a live campaign.

Common questions

1. 1.

   Do I need to pay for something extra to use the MCP server? No. The lemlist MCP server is available to all lemlist users. You just need an active lemlist account and an AI client that supports MCP.

2. 2.

   What’s the difference between OAuth and an API key in practice? OAuth is session-based. Tokens expire and refresh automatically, so there’s nothing to manage. An API key is permanent until you revoke it, which is more convenient for automated environments but requires careful handling to avoid credential leaks.

3. 3.

   My browser didn’t open during the OAuth flow. What should I do? Check that mcp-remote is installed and your terminal has permission to open URLs. You can also run npx mcp-remote https://app.lemlist.com/mcp manually to see the consent URL and open it yourself.

4. 4.

   Can I connect multiple lemlist teams? Yes. During the OAuth consent step, you select which team to authorize. To connect multiple teams, set up separate MCP server entries in your config, one per team.

5. 5.

   Which AI clients are supported? Officially documented clients are Claude Desktop, Claude Code, and Cursor. Any client that supports MCP over HTTP should work with the https://app.lemlist.com/mcp endpoint.

6. 6.

   Can I use lemlist MCP outside Claude and Cursor? Yes, and plenty of people already do. The endpoint is plain MCP over HTTP, so any client that speaks the protocol can call it. The most interesting pattern we see is CRM-embedded agents: Attio, for example, documents calling lemlist MCP tools inside a Custom agent workflow step, so a record update in the CRM can trigger lemlist actions without a human in the loop. Third-party builds exist too, including a Composio-built Claude Code Skill for lemlist, which is an unofficial path but a working one. The growing MCP ecosystem on GitHub has over 51,000 followers on the protocol’s organization page, and new clients ship constantly. If you’d rather use a no-code connection instead, our integrations page covers the non-MCP options.

You can also check the official API and MCP doc here.

Final thoughts

Connecting the MCP is the two-minute part. The payoff comes later, the first time you catch a dying sequence from a one-line question instead of a Monday morning dashboard crawl.

Start with OAuth, confirm one read-only call works (campaign stats is a good first test), then add the /lemlist-agent command and scope your buckets. From there you can run multichannel campaigns end to end without leaving your chat window.

Don’t have a lemlist account yet? Start a 14-day free trial. No credit card required, full access to all features, and the MCP server is included on every plan (lemlist is rated 4.6/5 on G2).