TL;DR: Customer.io launched its own official MCP server, and it changes what AI-assisted lifecycle marketing looks like at a fundamental level. Instead of describing your customer data to Claude and hoping it understands your setup, Claude now sees directly into your actual workspace; your real segments, campaign structures, message templates, and integration health.
The result is AI that gives specific, actionable recommendations grounded in your real data, not generic advice. This guide covers setup and the eight workflows that deliver the most immediate value.
Why the Customer.io MCP server changes the equation
Most marketers using AI for lifecycle work face the same problem: they have to explain their setup to the AI every single time. What your segments are called, how your journeys are structured, what attributes you track, what integration events you fire — none of that context travels with you into a generic AI chat window.
Customer.io's MCP server solves this at the root. When Claude is connected via MCP, it can query your workspace directly — reading segment definitions, listing campaigns, checking integration health, retrieving message delivery metrics — and use that live context to generate recommendations grounded in your actual setup.
Setting up the Customer.io MCP server
Step 1: Enable at account level. A Customer.io Account Admin must go to Settings → Privacy, Data & AI and enable both Customer.io AI and Customer.io MCP.
Step 2: Add to Claude Desktop.
- Go to Settings → Personal Settings → Connectors
- Select Add custom connector
- Enter the MCP server URL:
- US accounts:
https://mcp.customer.io/mcp - EU accounts:
https://mcp-eu.customer.io/mcp
- US accounts:
- Set transport type to HTTP
- Authenticate
⚠️ Data privacy reminder. Customer.io's MCP server returns information about real people in your workspace. Only connect AI tools your organisation has approved for processing customer data.
Workflow 1: Plain-text segment building
Customer.io's MCP server includes a segment builder that accepts plain-text descriptions and translates them into working segments. Claude can go further: checking your existing segments for overlap, suggesting refinements based on recent engagement data, and combining segments in ways that would require multiple filters to build manually.
Prompt:
Create a segment in Customer.io called "High-value dormant customers". Definition: customers who made at least 2 purchases in the last 12 months, have not made a purchase in the last 90 days, and have an AOV above £150. Before creating it, check our existing segments for overlap. Once created, tell me how many people are in the segment.Workflow 2: Data-grounded user persona development
Claude connected to your Customer.io workspace can query a segment, analyse behavioural patterns within it, and generate detailed, data-grounded personas that reflect what your users actually do — not what marketing assumes they do.
Prompt:
Using our "Power Users - Last 90 Days" segment in Customer.io, generate three distinct user personas. For each persona include: - Behavioural patterns (feature usage, engagement frequency, click patterns) - What distinguishes them from other personas - The lifecycle moment where they are most likely to churn or expand - The one message most relevant to them right now Base everything on actual attributes and events in this segment.Workflow 3: Campaign performance analysis
Ask Claude to pull your campaign performance from Customer.io, identify what is working and what is not, and produce a prioritised list of recommendations. Because Claude can see your campaign structure, segment definitions, and message templates alongside the performance data, recommendations are specific to your setup.
Prompt:
Pull delivery metrics for all campaigns sent in the last 30 days from Customer.io. For each campaign: sent volume, open rate, click rate, unsubscribe rate. Then: 1. Identify the 3 best-performing campaigns and explain why they worked 2. Flag the 3 worst-performing and diagnose the likely cause 3. Give me 5 specific changes to make to the bottom performers Use brand-voice.md when suggesting copy changes.Workflow 4: SDK integration and troubleshooting (Claude Code)
The Customer.io MCP server includes an integration tool specifically for developers setting up or troubleshooting SDK integrations — mobile SDKs, Node.js, or web-based. Describe the problem in plain English and Claude queries your workspace configuration, checks your SDK setup, verifies credentials, and returns specific diagnostic steps and code changes.
Prompt:
My users aren't receiving push notifications on Android for our trial expiry campaign. Check my Customer.io workspace: 1. Verify the Android source configuration 2. Check whether test device tokens are registered 3. Review push notification settings for the trial-expiry campaign 4. Identify any delivery errors in recent sends Return a diagnostic report with specific steps to fix each issue, and the code change needed to verify device token registration.Workflow 5: LLM Actions inside Customer.io campaigns
Customer.io's LLM Actions feature allows you to embed AI-powered steps directly inside your campaign workflows — sending data through a language model as part of a journey and storing the result as a journey attribute for use in subsequent messages.
Example LLM Action setup:
Trigger: User completes their 7th session in the app Input data: user.plan_type, user.most_used_feature, user.last_session_duration LLM task: Based on this user's usage pattern, write a 60-word personalised message recommending the one feature they haven't used yet that is most likely to increase their engagement. Be specific to their usage pattern. Store result as: journey_attribute.personalised_feature_recommendation Use in: next email send as Note: LLM Actions use AI credits. Test with a small cohort before rolling out to your full user base.
Workflow 6: Retention and re-engagement strategy
Ask Claude to look at your actual churned users, identify the behavioural signals that preceded their churn, compare them to retained users, and design a re-engagement campaign using your real segments and message templates as building blocks. The output is a strategy you can implement directly in Customer.io — not a framework you have to translate.
Prompt:
In Customer.io, look at users who cancelled in the last 60 days. Compare their behaviour in the 30 days before cancellation to users who were active in the same period and did not cancel. Identify: 1. Top 3 behavioural signals that preceded churn 2. The last campaign they engaged with before cancelling 3. At what point in their lifecycle most churn occurs Then design a 3-email re-engagement sequence for users showing these early warning signals. Post the full brief to ~/Documents/retention-brief.md.Workflow 7: Recurring broadcast management (Cowork)
For marketing teams running regular broadcasts, a Cowork thread connected to Customer.io manages the recurring workflow end to end: drafts the newsletter from your content sources, checks the recipient segment for current size and unsubscribe spikes, and routes the draft to your review queue on a schedule.
Workflow 8: Batch segment operations and workspace audits
For technical marketing teams and growth engineers, Claude Code connected to the Customer.io MCP server enables workspace audits: all active campaigns flagged for low open rates, stale segments worth archiving, campaigns with overlapping audiences, unused transactional templates, and webhooks that have not fired recently.
Preparing your Customer.io workspace for best results
Use consistent naming conventions for attributes and events — either snake_case or camelCase throughout. Add descriptions to your tags: tags are visible to AI features and help Claude understand what groups of assets have in common. Clear, descriptive segment names and campaign titles significantly improve the quality of Claude's recommendations.
Frequently asked questions
How do I connect Claude to Customer.io?
A Customer.io Account Admin must first enable both Customer.io AI and Customer.io MCP under Settings → Privacy, Data & AI. Then in Claude Desktop, go to Settings → Connectors, add a custom connector, and enter https://mcp.customer.io/mcp (US) or https://mcp-eu.customer.io/mcp (EU). A paid Claude plan is required.
What are LLM Actions in Customer.io?
LLM Actions are Customer.io workflow steps that send data to a language model mid-journey and store the result as a journey attribute for use in subsequent messages. Common uses include real-time personalised recommendations, dynamic message personalisation, sentiment analysis, and translation. LLM Actions use AI credits.
Can Claude trigger campaigns or send emails in Customer.io?
Claude can trigger broadcasts via the MCP server's Trigger Broadcast tool. For live audience sends, always include a human review step before execution. The recommended Cowork approach is to prepare the broadcast-ready draft and confirm all settings, then require explicit human approval before executing the trigger command.
Do I need to be technical to use Claude with Customer.io?
No. Marketers can use it for segment building, campaign analysis, persona development, and retention strategy in Claude Chat or Cowork. Engineers get additional value from SDK troubleshooting and workspace audits via Claude Code. The same connection serves both personas.
Next steps
- See the full Claude stack: The Complete Claude AI Stack
- Set up recurring Cowork workflows: How GTM Teams Can Maximise Claude Cowork
- Connect to your CRM too: How to Use Claude with HubSpot
- Build the GTM skills underneath: 15 Top Claude Skills for GTM Teams
About the author
Paul Sullivan is the Founder of ARISE GTM and creator of the ARISE GTM Methodology®. He is the author of Go To Market Uncovered (Wiley, 2025) and host of the GTM Uncovered podcast.
- Based: Here East, Queen Elizabeth Olympic Park, London E15 2GW
- Customer.io accreditation: Certified Implementation Partner
- Work with ARISE GTM: arisegtm.com/agency-services
- Speak to Paul: arisegtm.com/contact-us
- Podcast: GTM Uncovered on Spotify
- YouTube: @gtmuncovered
Based on ARISE GTM's lifecycle marketing engagements (2024–2026). Current as of April 2026.
%20small.jpg)
