Creating and deploying custom connectors

Build, test, and deploy a custom data connector in watsonx using Connector forge. The process includes generating connector artifacts with an AI coding agent, validating connectivity and authentication, and deploying the connector through serverless, OpenShift, or self-hosted options before registering it for use in watsonx services.

Creating a connector

  1. In your watsonx environment, open Platform connections and select the Connector forge tab.
  2. In the Connector forge tab, click the Build a connector.
  3. Copy the command and use it to clone Connector forge repository to your local machine.
  4. Open the cloned repository in Claude Code or Bob. Switch the agent to Connector forge custom mode that is pre-configured in the repository.
  5. Specify the name of the data source that you want to connect to in the AI coding agent, for example: "Build a connector for Salesforce".

The agent prompts you for required inputs in multiple steps until all connector configuration parameters are collected. During this process, the agent generates the connector artifacts. The primary output is a JSON DSL file that defines the connector.

  1. Once deployment is complete, return to the Connector forge window and click the Refresh icon. Your new connector will appear in a Select a connector table.
  2. (Optional) Edit the name and description of the connector by using the Edit icon in the Name and Description tabs.

Testing a connector locally

Before you deploy the connector, validate it locally to ensure that it passes all required checks. The AI coding agent runs a set of checks and guides you through resolving any issues.

  • Connectivity: Confirms that the connector can reach the target data source.
  • Authentication: Confirms that your credentials are valid and accepted.
  • Data retrieval: Confirms that the connector can read data from the data source. Review the test results and follow any prompts from the agent to fix issues before you proceed to deployment.

Deploying a connector

Choose a deployment option based on your environment and requirements.

Serverless deployment (fully automated)

Use this option if you want a fully managed, automated deployment.

  • Requires an existing IBM Cloud or AWS account.
  • The AI coding agent automates the full deployment to a serverless runtime (IBM Code Engine or AWS Lambda).
  • Registering the connector is automated.

OpenShift deployment (automated with manual fallback available)

Use this option if you manage your own cluster and you want the connector to run in the same cluster as your Cloud Pak for Data installation.

  • Requires customer-managed OpenShift or Cloud Pak for Data cluster
  • The AI coding agent automates deployment to your cluster using the platform APIs.
  • Registering the connector is automated.

Self-hosted server deployment (manual with agent-provided instructions)

Use this option if your environment does not permit granting the agent access to infrastructure APIs.

  • Requires a customer-managed Docker or Kubernetes environment
  • The AI coding agent provides a deployment package with a Docker image and instructions.
  • You must register the connector manually.

Registering the connector in watsonx

Register the connector to use it with watsonx services. If you choose the automated deployment, the AI coding agent completes this step automatically. For self-hosted manual deployment, follow the steps:

  1. In your watsonx environment, open Platform connections.
  2. Select New connection.
  3. From the connection list, select your custom connector.
  4. Provide the required connection details and save.