Waiting for engine...
Skip to main content

Using Boomi Companion to build a Data Hub model with Claude

This topic walks you through building a Boomi Data Hub model using Boomi Companion — from your initial prompt to a deployed model populated with verified test data.

Prerequisites​

  • A Claude license with the bc-datahub plugin or boomi-datahub skill installed
  • Access to the Boomi Enterprise Platform with the Data Hub service enabled
  • A Hub Cloud is available in your account (for example, USA East Hub Cloud). Data Hub repositories are hosted on Boomi-managed Hub Clouds. You do not need a Boomi runtime.
  • A .env file in your project workspace with Boomi Platform credentials (BOOMI_USERNAME, BOOMI_API_TOKEN, BOOMI_ACCOUNT_ID, BOOMI_API_URL). If you do not have one yet, the agent can walk you through the setup.

To install the bc-datahub plugin, run:

/plugin marketplace add OfficialBoomi/boomi-companion
/plugin install bc-datahub@boomi-companion

Core workflow​

Step 1: Provide a prompt​

Open your AI agent with the Boomi Companion bc-datahub plugin installed and enter a prompt. You do not need to provide API documentation, the plugin already knows the Data Hub APIs.

A minimal prompt is sufficient. Such as:

Build me a Data Hub model for my customer account data. 
Use standard customer account fields. 
I would like to use a repeating array based on type for addresses.

Boomi Companion will verify platform connectivity, design the model, and create it in your Boomi Platform account. The model will be created in a Draft state, so nothing is locked in yet.

Step 2: Review the model​

View your model in the Boomi Platform (Services > Data Hub > Models). You can request updates from the agent if you need changes.

important

This is the most important checkpoint in the flow. This is where you can add or rename fields, adjust enumeration values, and rethink your match rule, before publishing. Once published, field types and repeatability of the address group lock permanently.

Step 3: Publish the model​

Ask the agent to publish the model:

Go ahead and publish the model.

Publishing freezes the schema and assigns a version number - version 1 for a new model. Structural changes after this point require publishing a new version.

Step 4: Create a repository​

Ask the agent to create a repository or use an existing one:

Create a repository for this model.

A repository is the runtime container where you store golden records. The agent lists Hub Clouds available to your account and creates the repository there.

Step 5: Deploy the model to the repository​

Ask the agent to deploy the model:

Deploy the model to the repository.

Deployment binds the published model to the repository as a "universe" and provisions storage on the Hub Cloud. The agent confirms completion after polling.

Step 6: Wire up the Repository API credentials​

note

Complete this step manually, as it involves a secret that you store in your .env file rather than in the conversation.

Golden-record writes and quarantine checks are handled through the repository's own API, which requires a separate credential from the platform API token.

  1. Go to Services > Data Hub > Repositories > your repository > Configure tab.
  2. Copy the Hub Authentication Token from this tab.
  3. Add the following to your workspace's .env file:

DATAHUB_REPO_URI=<repository base URL from the Configure tab, ends in /mdm>

DATAHUB_REPO_USERNAME=<account-id>. <token-id> from the Configure tab

DATAHUB_REPO_AUTH_TOKEN=<the Hub Authentication Token>

Notify the agent when you are done. It verifies credentials without reading the values and confirms the token targets the intended repository. This check is important because a token for a different repository on the same cluster also authenticates successfully.

Step 7: Run the source and load test data​

Ask the agent to load test data:

Populate the model with some test data.

The agent runs the source's initial-load lifecycle, upserts records, and confirms golden records with no quarantine errors.

Results​

  • A published Data Hub model with standard customer account fields and a type-keyed repeating address collection.
  • A repository on a Boomi-hosted Hub Cloud with the model deployed as a universe.
  • A contributing source in incremental-update mode, ready for ongoing upserts.
  • Verified golden records with a clean quarantine.

From here, you can test key-based addresses, update behavior by resending an account with a changed BILLING address, and confirming it updates in place. You can preview match-rule behavior with a match request, or ask the agent to bootstrap a REST client connection so you can read and write golden records from a Boomi integration process.

On this Page