lemlist MCP review: 70 tools, three database sizes, and a 30-day reconnect

By Jan Suski. I go deep on every review, checking every number against its source and digging up details other coverage misses, a habit from six years of SEO and link building for tech companies. Bylines at Surfer, Brand24, and Senuto. How the scoring works.

Published 15 October 2026

Sales intelligence

The lemlist MCP server connects an AI assistant to lemlist through the Model Context Protocol, the open standard that lets a chat client call outside software as tools. Point Claude, ChatGPT, Cursor or Codex at one endpoint and the assistant can search leads, set up a sequence, launch campaigns and read reply rates from the chat window, without logging into lemlist directly.

This review covers what the 70 documented tools do, how setup works in each AI client, what breaks after a month, and where the vendor's pages contradict each other. The platform has its own lemlist review; this page sticks to the Model Context Protocol layer.

lemlist has an affiliate relationship with this site. That's why the complaints get as much room as the features.

↑ FACT SHEETWhat the lemlist MCP server does

The full tools reference lists 70 lemlist tools across 8 categories: campaigns and sequences (13), leads and sourcing (8), enrichment (7), CRM (10), inbox and messaging (3), analytics (6), deliverability and infrastructure (12), and account and automation (11). Accessed 15 October 2026, the page says only "Updated this week," and tool availability can change.

Half of them write to accounts, leads or campaigns. Thirty-five tools carry a "requires your confirmation" flag, among them Create campaign, Set campaign sender, Start campaign, Send message / reply and Purchase domain. The vendor's usage guide says "Tools that send, delete, spend credits, or change campaigns require your confirmation before running," while reads run without a prompt.

That prompt is the human-in-the-loop control, and it's the last one. Once a rep approves Start campaign, the sequence sends through the same mailboxes, warm-up and sending limits as campaigns built by hand.

Add sequence step accepts an email, LinkedIn or call step, so an assistant can create a sequence with a LinkedIn touch between email follow ups. Jamie H., head of growth at a software company, praised that base layer in an April 30, 2024 Capterra review: "It's easy to create campaigns with both email and LinkedIn automations in a simple workflow."

Two tools sit outside the pattern. Save memory stores an ICP or tone preference the assistant reuses, which is how enrichment context reaches personalized messages. Call lemlist API gives access to any endpoint of the lemlist API without a dedicated tool, running reads at once and asking before writes.

For lead generation, Search B2B database queries what the reference calls 450M+ contacts, and the tools to manage leads cover import, update, delete and push to CRM. Search campaign leads filters leads already in a campaign, and the ten CRM tools search, create and update contacts, companies and lists. Seven enrichment tools find or verify emails and phone numbers; nine tools in total spend credits, and LinkedIn enrichment needs the Chrome extension connected.

↑ FACT SHEETHow to connect lemlist: OAuth, API key and the consent page

Every client reaches the MCP server at one endpoint, https://app.lemlist.com/mcp. The setup article recommends OAuth, which needs no API key, and keeps the API key for clients that can't finish an OAuth flow or for shared automated setups.

Authentication, per the developer docs, is OAuth 2.1 with PKCE and dynamic client registration. On the first tool call a browser consent screen opens, the user picks a lemlist team, and tokens refresh automatically after that. In the legacy Claude Desktop config, mcp-remote handles discovery, client registration, the token exchange and refresh.

Claude Code and Claude Desktop

Claude Code takes one command, claude mcp add --transport http lemlist https://app.lemlist.com/mcp, and its default scope is local to the project where it ran. On a shared .mcp.json each teammate authenticates individually. A server that seems to vanish usually lives in another project; claude mcp list shows it. For a lemlist API key, add --header "X-API-Key: YOUR_API_KEY".

To connect lemlist in Claude Desktop, use Add custom connector. The legacy mcp-remote route needs Node.js and passes the API key through an API_KEY variable, and the vendor tells most people to skip it.

ChatGPT web, the ChatGPT desktop app and Codex CLI

ChatGPT web needs developer mode, an admin who allows custom apps, and a Business, Enterprise or Edu plan for full MCP. Pro users can build apps with the Apps SDK, and Plus and Pro users get pointed to the ChatGPT desktop app. The Codex CLI, the Codex IDE extension and the desktop app share ~/.codex/config.toml, where a [mcp_servers.lemlist] server block sets the URL and the key header.

Cursor reads ~/.cursor/mcp.json. On Cursor Enterprise an admin may need to allowlist the server under Team Settings, MCP Configuration, before anyone can connect lemlist to your AI editor.

↑ FACT SHEETToken lifetimes and the plan gate on each AI client

An access token lasts 1 hour and a refresh token 30 days. A connection idle for a month, or one whose user left the team they originally authorized, fails and has to authorize access again on the consent page.

Reply.io, which sells a competing server, called the 30-day window "a poor fit for an agent that runs monthly" in its August 2026 lemlist MCP review. That's a rival's opinion; the 30-day figure is the vendor's.

The other gates belong to the AI client. Claude Free allows only one custom connector. On Claude Team and Enterprise an owner adds connectors, and members connect to what an owner set. Switching tools mid-project means redoing the OAuth flow in each.

The troubleshooting list reads like a map of those gates: a ChatGPT connector that broke when custom servers moved into developer mode, Codex showing zero tools until login finishes, a consent page stopped by a pop-up blocker. A health check at GET /mcp/health confirms the server is up. The devil's in the details, and most details sit client-side.

↑ FACT SHEETWhat lemlist publishes about its own MCP

The vendor's pages aren't on the same page. The product page FAQ says the server "exposes 40+ actions"; the help center lists 70. The product menu says "650M+ Lead Database," the hero says 600M+, and the FAQ says "Search leads across 450M+ contacts."

Its speed claims come straight from the horse's mouth, with no method attached. The product page says setup takes under 2 minutes and that AI agents cut launching campaigns from 45 minutes to 90 seconds each. Those 90 seconds measure how fast campaigns get built. Deliverability decides what happens to campaigns after Start.

The same FAQ says "Your API key stays local on your machine." In the API key setup the key sits in a local config file and travels in an X-API-Key header to https://app.lemlist.com/mcp on every request. A lemlist API key, generated under Settings, Team, Integrations, opens the whole account: ?bucket= trims which tools the server advertises, and call_api still reaches the rest. Scoping the API key sets the real limit.

↑ FACT SHEETPricing and the credits an AI assistant spends

The pricing page lists "lemlist MCP & CLI" on the Email plan, $55 a user a month billed annually or $69 monthly, with 50,000 emails a month, alongside CRM integrations and API access. Multichannel runs $87 or $109 and adds LinkedIn, calls and SMS steps to campaigns, with 5 senders per user. Enterprise is quote-only, and the trial runs 14 days.

Credits are the real meter. Lead data enrichment sells at $16 a month annually for 2,000 credits, and the tools reference says enrichment cost depends on type and plan without a per-tool price. An assistant asked to find verified emails for leads in upcoming campaigns warns first, then draws down that pool.

Elena S., a biotech marketing manager, called the platform "Very good, very complete and not as expensive as other prospection tools" on Capterra on February 7, 2025. An agency reviewer saw the per-seat math differently:

"They don't even offer a discount on their $60/seat rate until you have had at least 20 seats."

Capterra review, Wil S., Creative Director, Media Production, June 20, 2022

↑ FACT SHEETWhat the MCP inherits from the platform

The analytics tools read the dashboard's data, so the dashboard's accuracy is the MCP's accuracy. Get campaign stats returns opens, clicks, replies, bounces and unsubscribes for campaigns, and an assistant asked to analyze performance builds on those numbers.

A Capterra reviewer, a management consulting CEO on an Enterprise contract, wrote on May 14, 2026 that "the reporting is literally unusable." Kelly K., in telecom marketing, wrote on Capterra on January 7, 2025: "The reporting is confusing." Ask an assistant to analyze performance across campaigns and it inherits both problems.

Deliverability travels with all campaigns an agent starts. Wil S. described the throttling:

"I started sending 150-200 emails per day from each seat and before long it became 50 at most."

Capterra review, Wil S., Creative Director, Media Production, June 20, 2022

A software-company marketing manager wrote on Capterra on September 3, 2022: "I was shocked to see that most emails landed in spam." The sending policy says accounts above its bounce and spam thresholds "can face account suspension and in some cases, termination," thresholds it can change "without prior notice."

An agent told to create campaigns from 500 unverified leads and hit Start is skating on thin ice under that policy.

Support gets mixed marks. Chris P., a telecom marketing director, wrote on Capterra on September 28, 2023:

"But their real asset is the support and guidance, continuously providing ideas and ways to improve."

Capterra review, Chris P., telecom marketing director, September 28, 2023

Ben F., a two-year user in online media, wrote on Capterra on May 3, 2024 that "Their support is not technical and only relays what their technical people tell them."

Refunds raise another red flag. Cyrus M., an owner in marketing, wrote on Capterra on April 27, 2021 that "they refused to refund me for the unused subscription, which was a huge red flag to me."

↑ FACT SHEETlemlist MCP against Amplemarket, Seamless.AI, Lusha and Reply

Four sales outreach tools ship a comparable MCP server. Figures for the first three come from this site's own MCP reviews of each.

Amplemarket MCP exposes 63 named tools on plans starting at $600 a month billed annually. It wins for a team that wants signals, enrichment and leads in one data pool, and it loses on entry price against the $55 Email seat.

Seamless.AI's MCP counts 54 tools and connects on a free plan of 50 credits and one seat. A rep who wants to look up and enrich contacts picks Seamless. A team that wants an assistant to create and manage campaigns and send replies stays with lemlist.

Lusha MCP runs 22 tools on any Lusha plan, Free included. It looks up leads and contacts with zero campaign tools, and Lusha carries a pending class action covered in that review.

Reply.io argues its own MCP handles replies to inbox messages and LinkedIn step setup, prefers a scoped API key over OAuth, and charges $69 a month per connected LinkedIn account. The Send message / reply tool weakens the first claim. Reply wins for a team running campaigns that wants scoped keys from day one.

Two tools pair with the lemlist MCP server. Clay fills email and phone gaps with waterfall enrichment before leads reach campaigns, and the folk CRM MCP keeps relationship data where an assistant can reach it.

The MCP tools list covers the full roster, including Capsule CRM's MCP. For data, see the email finder tools ranking, the sales intelligence ranking and the sales engagement pricing report. Apollo and Woodpecker compete at the platform level.

↑ FACT SHEETVerdict

The lemlist MCP server fits a founder or small team already running outreach on the platform. It ships with the $55 Email plan, connects in one Claude Code command, and covers 70 lemlist tools that search leads, create campaigns, enrich contacts and check deliverability, 35 of them behind a confirmation prompt.

It doesn't fit an agency running unattended agents across client accounts. Each OAuth consent binds one lemlist team, refresh tokens lapse after 30 idle days, and ChatGPT web needs a Business plan or above. Anyone picking the platform for the MCP alone should read the reporting and deliverability complaints about live campaigns first.

A mid-sized team can make it work with three habits: set a scoped API key, because ?bucket= trims the menu and leaves call_api open; check stats on live campaigns against a second source before trusting an assistant on reply rates; and keep an eye on the credit pool, since nine tools draw from it.