Create a custom connector from scratch

To create a custom connector, you must describe the API you want to connect to so that the connector understands the API's operations and data structures. In this article, you create a custom connector from scratch, without using an OpenAPI definition to describe the Azure Cognitive Services Text Analytics API sentiment operation (our example for this series). Instead, you describe the connector completely in the custom connector wizard.

For another way to describe an API, go to Create a custom connector from an OpenAPI definition.

Prerequisites

• An API key for the Cognitive Services Text Analytics API

• One of the following subscriptions:

  • Power Automate
  • Power Apps

Start the custom connector wizard

1. Sign in to Power Apps or Power Automate.

2. On the left pane, select Solutions.

3. Edit or create an unmanaged solution that will include your custom connector. Learn how to create a solution

4. Select the New drop down and select Automation > Custom connector.

   The new connector wizard opens

5. Type the Connector Name, such as SentimentDemo.

   Complete the rest of the steps using the following instructions.

Step 1: Update general details

From this point, we show the Power Automate UI, but the steps are largely the same across the technologies. We point out any differences.

On the General tab, do the following:

1. In the Description field, enter a meaningful value. This description appears in the custom connector's details, and it can help others decide whether the connector might be useful to them.

2. Update the Host field to the address for the Text Analytics API. The connector uses the API host and the base URL to determine how to call the API.

   Parameter	Value
   Description	Uses the Cognitive Services Text Analytics Sentiment API to determine whether text is positive or negative
   Host	westus.api.cognitive.microsoft.com

Step 2: Specify authentication type

There are several options available for authentication in custom connectors. The Cognitive Services APIs use API key authentication, so that's what you specify for this tutorial.

1. On the Security tab, under Authentication type, select API Key.

2. Under API Key, specify a parameter label, name, and ___location. Specify a meaningful label, because this is displayed when someone first makes a connection with the custom connector. The parameter name and ___location must match what the API expects.

   Parameter	Value
   Parameter label	API key
   Parameter name	Ocp-Apim-Subscription-Key
   Parameter ___location	Header

3. At the top of the wizard, make sure the name is set to SentimentDemo, and then select Create connector.

Step 3: Create the connector definition

The custom connector wizard gives you many options for defining how your connector functions, and how it's exposed in logic apps, flows, and apps. We explain the UI and cover a few options in this section, but we also encourage you to explore on your own.

Create an action

The first thing to do is create an action that calls the Text Analytics API sentiment operation. On the Definition tab, the left pane displays any actions, triggers (for Logic Apps, Power Automate, and Copilot Studio), and references that are defined for the connector.

1. Select New action.

2. In the General area, add a summary, description, and operation ID for this action.

   Parameter	Value
   Summary	Returns a numeric score representing the sentiment detected
   Description	The API returns a numeric score between 0 and 1. Scores close to 1 indicate positive sentiment, while scores close to 0 indicate negative sentiment.
   Operation ID	DetectSentiment

   Leave the Visibility property set to none. This property for operations and parameters in a logic app or flow has the following options:

   • none: Displayed normally in the logic app or flow
   • advanced: Hidden under another menu
   • internal: Hidden from the user
   • important: Always shown to the user first

3. In the Request area select Import from sample.

4. Specify the information necessary to connect to the API, specify the request body (provided after the table), and then select Import.

   We provide this information for you, but for a public API, you typically get this information from documentation such as Text Analytics API (v2.0).

   Parameter	Value
   Verb	POST
   URL	https://westus.api.cognitive.microsoft.com/text/analytics/v2.0/sentiment
   Body	Use the example JSON.

   Example:

   {
     "documents": [
       {
         "language": "string",
         "id": "string",
         "text": "string"
       }
     ]
   }
   

5. In the Response area, select Add default response.

6. Specify the response body, and then select Import. As we did for the request body, we provide this information for you, but it's typically provided in the API documentation.

   Example:

   {
    "documents": [
      {
        "score": 0.0,
        "id": "string"
      }
    ],
    "errors": [
      {
        "id": "string",
        "message": "string"
      }
    ]
   }
   

   The Validation area displays any issues that are detected in the API definition.

7. Fix any issues. You should see a green check mark when the definition validation succeeds.

8. In the upper-right corner of wizard, select Update connector.

Update the definition

Let's change a few things so that the connector is more friendly when someone uses it in Logic Apps, Power Automate, Power Apps or Copilot Studio.

1. In the Request area, select body, and then select Edit.

2. In the Parameter area, you now see the three parameters that the API expects: id, language, and text. Select id, and then select Edit.

3. In the Schema Property area, update values for the parameter, and then select Back.

   Parameter	Value
   Title	ID
   Description	An identifier for each document that you submit
   Default value	1
   Is required	Yes

4. In the Parameter area, select language > Edit, and then repeat the process that you used for id to add the following language values:

   Parameter	Value
   Title	Language
   Description	The two or four character language code for the text
   Default value	en
   Is required	Yes

5. In the Parameter area, select text > Edit, and then repeat the process that you used for id and language to add the following text values:

   Parameter	Value
   Title	Text
   Description	The text to analyze for sentiment
   Default value	None
   Is required	Yes

6. In the Parameter area, select Back to return to the main Definition tab.

7. In the upper-right corner of the wizard, select Update connector.

Step 4: (Optional) Use custom code support

Custom code transforms request and response payloads beyond the scope of existing policy templates. Transformations include sending external requests to fetch additional data. When code is used, it takes precedence over the codeless definition. This means that the code will execute, and we don't send the request to the back end.

You can either paste in your code or upload a file with your code. Your code must:

• Be written in C#.
• Have a maximum execution time of five seconds.
• Have a file size no larger than 1 MB.

For instructions and samples of writing code, go to Write code in custom connectors.

For frequently asked questions about custom code, go to Custom code FAQ.

1. On the Code tab, insert your custom code by using one of the following options:

   • Copy/paste
   • Select the Upload button.

   If you choose to upload your custom code, only files with a .cs or .csx extension are available.

   

   Important

   Currently, we only support syntax highlighting in the code editor. Make sure to test your code locally.

2. After you paste or upload your code, select the toggle next to Code Disabled to enable your code. The toggle name changes to Code Enabled.

   You can enable or disable your code anytime. If the toggle is Code Disabled, your code is deleted.

3. Select the actions and triggers to apply to your custom code by selecting an option in the dropdown menu. If no operation is selected, the actions and triggers are applied to all operations.

   

Step 5: Test the connector

Now that you created the connector, test it to make sure it's working properly. Testing is currently available only in Power Automate and Power Apps.

1. On the Test tab, select New connection.

2. Enter the API key from the Text Analytics API, and then select Create connection.

   Note

   For APIs that require bearer authentication, add Bearer and one space before the API key.

3. Return to the Test tab, and do one of the following:

   • (In Power Automate) You're taken back to the Test tab. Select the refresh icon to make sure the connection information is updated.

     

   • (In Power Apps) You're taken to the list of connections available in the current environment. On the left pane, select Custom connectors. Choose the connector you created, and then go back to the Test tab.

4. On the Test tab, enter a value for the text field (the other fields use the defaults that you set earlier), and then select Test operation.

   The connector calls the API.

5. Review the response, which includes the sentiment score.

   

Best practices for CLI users

• Download all your connectors, and use Git or any source code management system to save the files.

• If there's an incorrect update, redeploy the connector by rerunning the update command with the correct set of files from the source code management system.

• Test the custom connector and the settings file in a test environment before deploying in the production environment.

• Always double-check that the environment and connector ID are correct.

Next steps

Now that you created a custom connector and defined its behaviors, you can use the connector from:

• Power Automate
• Power Apps
• Logic Apps
• Copilot Studio

You can also share a connector within your organization, or get the connector certified so that people outside your organization can use it.

• Share your connector
• Certify your connector

Provide feedback

We greatly appreciate feedback on issues with our connector platform, or new feature ideas. To provide feedback, go to Submit issues or get help with connectors and select your feedback type.