What is Loomi Connect?

Loomi Connect is Bloomreach's interface layer for AI agents. It exposes Bloomreach's commerce intelligence, marketing automation, analytics, search, merchandising, and conversational capabilities as structured tools that AI agents and applications can discover and use.

With Loomi Connect, you can build agentic experiences that connect Bloomreach intelligence to external systems, partner platforms, and custom agent frameworks. Loomi Connect supports multiple integration methods, so you can connect using the approach that fits your stack.

🚧
Early access disclaimer
This integration is a technical preview and the concept is under development. The interface and availability are subject to change.

What agents can do with Loomi Connect

Generic AI agents usually run on shallow context, such as a few fields from a CRM, a product feed, or a reporting API. Loomi Connect grounds agents in your full Bloomreach setup, so they can act on real customer profiles, product data, journeys, and analytics instead of stitching together brittle integrations.

An agentic workflow powered by Loomi Connect can:

Fetch rich customer and product context
Gather multi-channel context across email, SMS, push, web, ads, search
Make real-time decisions (offers, rankings, journeys, analytics)
Automate Bloomreach workflows

Loomi Connect use cases

Loomi Connect provides live Bloomreach data to agents, supporting use cases across product discovery, customer lifecycle, and analytics. The patterns below are common starting points:

Conversational product discovery

Agents translate natural language into search queries, filters, and bundles across web, app, and emerging LLM surfaces. Shoppers can ask for things like "a gift under $50 that ships this week" and get relevant results without navigating menus or filter trees. The agent draws on your product catalog, recommendations, and customer context to interpret intent and shape responses.

Churn signal detection

Agents detect churn patterns across visits, preferences, and engagement — beyond simple "days since last purchase" rules. By querying customer events, segmentations, and prediction scores together, they surface customers at risk based on behavioral signals, not just recency. Lifecycle teams get a richer view of who's likely to disengage and why.

Agentic analytics and benchmarking

Agents query performance metrics like revenue per visitor, customer lifetime value, and journey drop-offs. They pull from saved analyses, dashboards, and ad-hoc queries to answer "what happened" and "what should we try next" in the same conversation. Useful for QBR prep, deal reviews, and post-campaign retros.

Cross-channel performance views

Agents pull from email, SMS, ads, search, and on-site behavior into a unified view of performance. They surface anomalies — like a segment's open rate spiking — and opportunities, like products over-performing in search impressions but under-converting. Useful when you'd rather ask a question than build another dashboard.

How Loomi Connect works

Loomi Connect exposes Bloomreach capabilities as structured tools that AI agents can discover and invoke.

Connect via MCP

MCP is the only integration method supported right now. Loomi Connect runs as a remote MCP server over Streamable HTTP transport, so there's nothing to install locally. To use Loomi Connect through MCP, you take three steps:

Connect the MCP server to your AI application. Loomi Connect works with any MCP-compatible AI application, agent, or development environment, including Claude Code, Cursor, VS Code, and Windsurf. Point your client at the Loomi Connect server URL. No API keys or custom headers are needed.
Authenticate the connection. Loomi Connect uses your existing Bloomreach login through SSO (OIDC), so you don't need separate credentials. A browser window opens the first time your assistant calls a tool, and your session persists for up to 30 days. Your AI application sees exactly the data your account has permission to access, respecting your IAM permissions.
Work with MCP capabilities. Once authenticated, your AI application can call Loomi Connect tools to read live data from your Bloomreach workspace. Tools follow the organization > workspace > project hierarchy, so most calls require a project ID.

For setup instructions on different platforms, see Get started with MCP.

Available capabilities

Loomi Connect provides read-only access to Engagement and Discovery data. It can't create, edit, or delete anything in your workspace. Your AI application can explore:

Inspect customer data, including identifier schemas, properties, events, and prediction scores.
Review campaigns, automation scenarios, and scheduled activity across channels.
Explore audiences, segmentations, customer filters, and autosegments.
Query analytics, including funnels, trends, reports, and dashboards.
Inspect audiences, analytics, recommendation engines, and their dependencies.
Query product catalogs, A/B test data, and product grid insights.
Review rule configurations for search, categories, autosuggest, and other merchandising operations.
Audit workspace or project settings, including consent, frequency policies, and voucher pools.
Support multi-step workflows across Bloomreach and partner ecosystems.

📘
Note
Loomi Connect also serves built-in workflow guides as MCP resources. These load automatically into your AI assistant's context and explain how to navigate the project hierarchy, look up customers or products, explore scenarios or campaigns, and run ad-hoc analytics queries.

Limitations

Loomi Connect is in early access. Keep the following constraints in mind when building on top of it.

Scope

Read-only tools: Creating, editing, or deleting campaigns, segments, customers, or any other object isn't supported in this phase. Write operations, bulk export, or webhook delivery aren't available at the moment.

Data access

PII masking: Customer property values follow your IAM permissions. Users without the required permissions see masked values (******) for PII fields.
365-day event window: Customer event history is limited to the last 365 days. There's no way to request events outside this window.

Authentication

Session length: Authenticated sessions last up to 30 days. After expiry, the next tool call triggers a browser re-authentication prompt.
Multiple accounts: If you have multiple Bloomreach accounts, the first successful login is used for the session. Switching accounts requires re-authentication.

Performance and rate limits

Refer to the MCP performance and rate limits guide for details on performance expectations, and the behavior of rate limiting and caching.