Get started with MCP

This guide walks you through connecting an MCP-compatible AI application or platform to the Loomi Connect MCP servers. Once connected, your AI application can read live data from your Bloomreach workspace through more than 160 tools spanning marketing, analytics, search, and Data hub, including write tools that create, update, and delete configuration.

📘
Note
Loomi Connect is currently in Beta. It is available by default to all orgs on the unified login.
Join the Loomi Connect Slack channel — it's the main place for updates, support, and the occasional reminder that the platform changed again in the last 48 hours.

Prerequisites

A Bloomreach account with access to the workspace you want to query. The MCP server respects your existing Bloomreach IAM permissions, including PII masking.
One of the following AI applications, or another platform that supports MCP connections:

Setup overview

Loomi Connect runs remote MCP servers, so there's nothing to install locally. Connect to an AI application in three steps:

Add a Loomi Connect server URL to your AI application's MCP configuration.
Restart or reload the AI application so it picks up the new server.
Trigger a tool call to start the browser-based authentication flow.
Test MCP capabilities in your AI application.

Connect the MCP server

⚠️
Important
Connect to one or more region-specific MCP servers that match your project or account configuration. Partner organizations can build demos using the demo server at http://brx.connect.loomi.ai/mcp.
Contact [email protected] if you face issues with the connection or access.

Follow the steps for your preferred environment to connect the MCP server:

Run one or more of these commands in your terminal:

claude mcp add loomi-connect-usa --transport http https://us.connect.loomi.ai/mcp

claude mcp add loomi-connect-europe --transport http https://eu.connect.loomi.ai/mcp

claude mcp add loomi-connect-uk --transport http https://uk.connect.loomi.ai/mcp

claude mcp add loomi-connect-canada --transport http https://ca.connect.loomi.ai/mcp

claude mcp add loomi-connect-apac --transport http https://ap.connect.loomi.ai/mcp

To verify the server was added, run:

claude mcp list

Refer to the Claude Code documentation for more details on MCP configuration.

Custom Python MCP clients

If you build your own client with the official mcp Python SDK against a Loomi Connect URL (https://eu.connect.loomi.ai/mcp and the other regional /mcp endpoints), pin the following in your dependencies:

mcp>=1.9,<2

📘
Note
Claude Code, Cursor, VS Code, and Windsurf do not need a Python package pin. They use a built-in MCP HTTP client.

Unpinned installs resolve to mcp 2.x, which removes streamablehttp_client and fails with:

ImportError: cannot import name 'streamablehttp_client'

If you see this ImportError, pin mcp>=1.9,<2 and retry. Do not treat it as a Loomi Connect outage.

🚧
Important
The MCP SDK 2.x is a code migration, not a drop-in upgrade. It renamed the helper to streamable_http_client and changed HTTP client / OAuth types. See the MCP Python SDK v1 to v2 migration guide. Loomi Connect has not certified SDK 2.x yet.

Verify the connection

After restarting your AI application, confirm that Loomi Connect is registered and available in the project:

In Claude Code, type /mcp in a new session.
In Cursor, open the MCP settings panel (File > Preferences > Cursor Settings > MCP).
In VS Code, open the Extensions tab to see installed MCP servers. You can also type @mcp in the command palette.

The first time your AI application calls a Loomi Connect tool, a browser window opens to the Bloomreach login page so you can authenticate. Sessions persist for up to 30 days.

Next steps

Authenticate MCP connection: sign in with your Bloomreach account and confirm your AI application can read live workspace data.