Configuring Claude Desktop integration
You can configure the Claude Desktop integration in HCL DevOps Loop to enable Loop Genie services by connecting to the Claude Desktop provider.
Before you begin
You must have completed the following tasks:
- Ensured that you have administrative privileges in DevOps Loop.
- Completed the steps to prepare your Claude Desktop account for integration with DevOps Loop. See Prerequisites for setting up AI providers for Loop Genie.
Procedure
-
Click .
The Integrations page is displayed.
-
Click
.
The New Integration pane is displayed with the Overview tab. - Enter a name to identify the integration in Name.
- Select Claude from the Provider list.
- Select one of the following options in the Credential Type list:
- Enter the API key of your Claude account to authenticate the connection in API Key and go to step 8.
-
Enter the details in the following fields:
- Username
- Password
- Confirm Password
-
Click Next.
The Configure tab is displayed.
-
Perform the following steps in the Configure tab to set
the parameters:
- Enter the API Endpoint URL used by your Claude account.
- Select Chat in Mode to set the mode of communication.
- Select an LLM to be used from the Model list.
- Enter the Organization ID configured in your Claude account for your organization.
- Enter the Project ID configured in your Claude account for your organization.
- Enter one or more Stop Sequences to define where the model should
stop generating text. For example:
"stop"or"end". - Use the Temperature slider to control the creativity of the
response:
-
Lower values produce more precise, consistent answers.
-
Higher values produce more creative or varied output.
-
- Set the Max Length slider to define the maximum number of tokens the model can generate.
- Use the Top P slider to control the diversity of
the generated output:
-
Lower values focus on the most likely words.
-
Higher values allow more varied responses.
-
- Set the Frequency Penalty slider to reduce repetition of words or
phrases.
- Higher values result in less repetitive output.
- Set the Presence Penalty slider to encourage the model to introduce new topics rather than repeating previous ones.
- Set the Best Of slider to choose how many internal responses the model generates before returning the best one.
- Click Save to apply the configuration.
