Customer.io MCP Server

The Customer.io MCP server is an officially hosted remote MCP endpoint that gives AI assistants direct access to a Customer.io workspace. Instead of describing customer attributes, segments, or campaigns to an AI, the assistant can introspect the workspace itself: discover available API endpoints, read profiles, generate segments grounded in real attributes and behavior, and (with opt-in) make write requests against the Journeys UI API and CDP Data Pipelines API.

The server exposes a small, composable set of tools rather than one tool per API endpoint. The core pattern is cio_prime and cio_schema to load the AI-ready API reference and discover endpoints, followed by cio_api to actually execute calls. A skills system (cio_skills_list, cio_skills_read) provides multi-step playbooks for common marketer workflows. This design keeps the tool surface small while covering the full Customer.io API.

It is a provider-hosted remote server reached over HTTP, with OAuth authentication tied to each user's Customer.io login. There is no GitHub repo to clone or local install to run. Customer.io operates separate endpoints for the US and EU regions, and the feature must be enabled at the account level under AI settings.

Tools

Prerequisites

• A Customer.io account with admin access
• An account admin must enable both Customer.io AI and Customer.io MCP under Settings > AI
• An MCP-compatible client (Claude, ChatGPT, Cursor, etc.). Note: MCP support is not available on Claude's free plan.

Server URL

Use the endpoint that matches your workspace region:

• US: https://mcp.customer.io/mcp
• EU: https://mcp-eu.customer.io/mcp

Transport type must be http. Server-sent events (sse) are not supported.

Claude Desktop setup

1. Open Settings > Connectors.
2. Enable the built-in Customer.io connector, or add a custom connector with the server URL above.
3. Click Connect and authenticate with your Customer.io credentials via OAuth.
4. Select the workspaces you want to expose.

Generic MCP client config

For clients that use a JSON config block, point a remote MCP entry at the URL:

{
  "mcpServers": {
    "customer-io": {
      "type": "http",
      "url": "https://mcp.customer.io/mcp"
    }
  }
}

You can review and revoke active sessions under your Customer.io personal settings in the Connected clients section.

• Generate or refine customer segments grounded in the actual attributes and events in your workspace, rather than generic examples.
• Inspect an individual customer profile to understand their journey state, attributes, and recent events while debugging a campaign.
• Search the workspace for existing campaigns, segments, broadcasts, or messages by name or tag before creating duplicates.
• Use cio_api to script ad hoc reads against the Journeys UI API or CDP Data Pipelines API for one-off analyses.
• Run pre-built skills (via cio_skills_list / cio_skills_read) to execute multi-step marketer workflows like onboarding audits or list hygiene.

• "Show me the schema for segments and then list the 10 most recently created segments in this workspace."
• "Look up the profile for user_id 48219 and summarize their last 5 events and which campaigns they are in."
• "Propose three new segment definitions for high-intent trial users based on attributes that actually exist in the workspace."
• "Find all active campaigns tagged onboarding and tell me which ones have not been edited in the last 90 days."
• "List available Customer.io skills and run the one for auditing unsubscribe handling."

• Officially built and hosted by Customer.io, with OAuth login and per-user authentication.
• Compact tool surface (six tools) that covers the full Journeys UI and CDP Data Pipelines APIs through cio_api.
• Write operations are off by default, reducing the risk of an agent making unintended changes.
• Region-aware endpoints for US and EU workspaces.

• Requires an account admin to enable Customer.io AI and MCP at the account level before any user can connect.
• Not usable on Claude's free plan, and sse transport is not supported.
• Generic tool names like cio_api mean the model must first call cio_prime and cio_schema to be effective, which adds round trips compared to servers that expose one tool per endpoint.

• Customer.io CLI (terminal-based AI integration from Customer.io for similar workflows).
• Pipedream-hosted Customer.io MCP, which proxies the Customer.io API through Pipedream's MCP layer.
• Zapier MCP with Customer.io actions, for users who prefer Zapier's auth and action catalog.