Rémi KokabiNovember 4, 20267 min read

Amplitude integration with lemlist through MCP, with prompts to copy

A free workspace hits three active users in a single week. Someone on that team is testing hard, maybe ready to buy, maybe about to churn if nobody shows up. The data sits in Amplitude. The outreach should start in lemlist. Instead the two tools sit on separate tabs, and by the time a human notices the spike, the moment has passed.

This is what the amplitude lemlist integration through MCP fixes. Connect both MCP servers in the same AI client and an agent reads usage data straight from Amplitude, then builds and launches the right campaign in lemlist, no export, no manual list-building, no RevOps ticket.

In this article, I’ll show you exactly why this connection matters and what you need before you start. Then I’ll give you three prompts you can copy right now: one for sales assist on active free accounts, one for catching usage drops before churn, and one for following up on key feature adoption. Each comes with what the agent actually does behind the scenes and what to check before you let it run on a full list.

Product usage is the clearest signal a SaaS company has. This is how you stop leaving it sitting in a dashboard nobody checks.

Why connect lemlist and Amplitude through MCP

Product usage data is the strongest buying signal a SaaS company owns, and most of it sits inside Amplitude. When both MCP servers are connected in the same AI client, an agent can read that usage data and start the right outreach in lemlist without a CSV export, a Zapier step, or a data pipeline in between.

An MCP server is a lightweight layer that exposes a tool’s actions to an AI agent through a standard protocol. The agent discovers what it can do (search leads, query charts, create campaigns) and calls those actions during a conversation. If you want the full picture of how this works for outbound, read the guide on MCP for sales.

Because the Amplitude MCP lets the agent query product analytics and the lemlist MCP lets it search contacts, enrich them, and build sequences, the two together turn a usage event into a live campaign in a single prompt.

What you need before you start

  • The lemlist MCP connected in Claude, ChatGPT, or Cursor. Follow the steps in how to set up the lemlist MCP if you haven’t done this yet.

  • The Amplitude MCP connected in the same client, using your Amplitude API key or OAuth credentials and the server URL for your region.

  • Permissions scoped to what the use cases below actually need. On the Amplitude side, a read-only role is enough for every prompt here. On the lemlist side, campaign creation and lead management permissions cover everything.

Start with read-only prompts that pull data from both tools before you let the agent write in either one. That way you see exactly what it plans to do before anything changes in your campaigns or analytics.

1. Sales assist on active free accounts

Free workspaces that adopt fast rarely hear from sales. By the time someone notices, those users have already formed habits, hit limits, and decided on their own whether to upgrade or leave.

Find free workspaces in Amplitude with 3 or more active users this week, get the decision maker at each company from lemlist, and add them to the sales assist campaign.

The agent queries Amplitude for workspaces on a free plan with three or more active users in the current week. It then searches the lemlist People database for the decision maker at each company, pulling verified contact details through waterfall enrichment. Finally, it adds each lead to the sales assist campaign and passes the usage count as a custom variable so the first message can reference how active the team already is.

Before running this on a full list, confirm the campaign exists, the sequence is reviewed, and the custom variable name matches what your template expects.

2. Save accounts whose usage drops

Churn shows up in product data weeks before the cancellation email. A paying account whose weekly active users fall by half is telling you something, and waiting for a support ticket means waiting too long.

List paying accounts in Amplitude whose weekly active users dropped by half this month and start a check-in sequence for their admins.

The agent reads the weekly active user trend for paying accounts in Amplitude and filters for those where the count dropped by 50% or more within the current month. It then finds the admin contacts for each account in lemlist and adds them to a check-in campaign. The sequence can open with a line that acknowledges the team has been less active, without sounding like surveillance, because the variable carries the trend rather than raw numbers.

Speed to lead matters here as much as it does on inbound. According to lemlist internal data, leads contacted the same day as a signal reply at 21.3%, while those contacted after a week reply at 8.1%. Check that the campaign tone fits a retention conversation, and test on a batch of 10 accounts before opening it up.

3. Follow up on a key feature adoption

Users who adopt a key feature for the first time are telling you they need more from the product. That moment is the best opening for an upgrade conversation because the value is fresh and the context is specific.

Find users in Amplitude who used feature X for the first time this week and add their team lead to the upgrade campaign.

The agent queries Amplitude for the first occurrence of the feature event within the current week. It identifies the company tied to each user, searches the lemlist People database for the team lead or manager at that company, and adds them to the upgrade campaign with the feature name and first-use date as custom variables.

Before letting this run on a full list, replace “feature X” with the actual event name in your Amplitude taxonomy and verify the upgrade campaign sequence references the right variable.

Guardrails when an agent works in lemlist and Amplitude

Giving an agent write access to two tools at once is powerful, but it needs boundaries. Follow these before you scale any of the prompts above.

  • Review every campaign the agent creates or updates before you launch it. The agent can build the sequence, but a human should approve the messaging and the audience.

  • Test on 10 leads first. Run the prompt with a small batch, check the leads that were added, and confirm the variables populated correctly.

  • Keep unsubscribes and exclusion lists active in lemlist. The agent respects these lists when adding leads, but only if they are turned on.

  • Limit who on the team can connect the MCPs. One or two people should own the API keys and the client configuration so you have a clear audit trail.

  • Log what the agent changed. After each session, ask the agent to summarize the leads it added, the campaigns it touched, and any variables it set. Save that summary for your records.

Run it in lemlist

The lemlist MCP server works in Claude, ChatGPT, and Cursor. Once connected, the agent can do everything you would do inside lemlist, just from a prompt.

It can search the People database (650M+ contacts, 65M+ companies), enrich leads with waterfall enrichment across 14+ providers, and create campaigns and sequences across email, LinkedIn, calls, and WhatsApp. It reads the unified inbox and campaign stats so you can ask follow-up questions about performance without switching tabs. And it can act on intent signals, turning a product event in Amplitude into a live outreach step.

For a full walkthrough of what the agent can do on its own, see the lemlist MCP use cases guide.

Final takeaway

Product usage data in Amplitude and outreach tools in lemlist belong in the same conversation. With both MCP servers connected, an agent turns adoption spikes, usage drops, and feature events into campaigns without a single export or manual step. The prompts above are ready to copy, test on a small batch, and scale.

Build your first signal-based sequence in lemlist.

Launch AI outbound campaigns

Automate prospecting and personalize AI-powered sequences to book more meetings.

Frequently asked questions

Through MCP, yes. Connect both MCP servers in the same AI client and the agent can read Amplitude data and act in lemlist in one conversation.

Any client that supports remote MCP servers, such as Claude, ChatGPT, or Cursor, with both servers connected.

It can create and update campaigns, but review each campaign before launch and test on a small batch first.

No. You connect both MCP servers in your AI client, then write your requests in plain language.
Rémi Kokabi

Rémi Kokabi

Co-founder @GTM Club

Share this post