Custom API Integration allows you to connect SleekFlow AI modules, such as AgentFlow, with external systems using API details you provide.

With a Custom API integration, your AI agent can retrieve external data, trigger actions, or update records during customer conversations. This helps your agent support workflows that require real-time business data instead of relying only on static Knowledge Base responses.

For example, your agent can:

• Create payment links through your internal system
• Retrieve customer or membership details from a CRM
• Search products or generate booking links
• Trigger backend workflows based on customer requests

You can set up a Custom API integration by describing your API use case, uploading supported API documentation, pasting relevant API documentation into the setup chat, or pasting a raw cURL command into the setup chat. The Custom API builder uses the information you provide to generate an editable schema.

Before you begin

Before setting up a Custom API integration, make sure you have:

• Access to the Integrations tab in SleekFlow AI
• A clear understanding of what you want the AI agent to do with the API
• API documentation, endpoint details, or request details for the endpoint you want to connect
• API key, bearer token, or authentication header details, if your API requires authorization

If the Integrations tab is unavailable or locked, contact your Customer Success Manager to enable or upgrade access.

You should also prepare the API information needed for setup. This can include:

• A supported API documentation file
• Relevant API documentation text
• A raw cURL command
• A short description of the API use case

Supported upload formats include:

• .JSON
• .YAML
• .YML
• .PDF

Each uploaded file can be up to 1 MB.

If Copilot suggests an endpoint or generates a schema for you, always review the details before saving the integration. Check the endpoint, Base URL, authentication headers, inputs, outputs, and field descriptions to make sure they match your API requirements.

How Custom API Integration works

Custom API Integration defines the API schema that AgentFlow can use.

This includes:

• Inputs: Data sent to the API
• Outputs: Data returned from the API
• Descriptions: Information that helps the agent understand what each field means
• Data source: Where each input value comes from
• Base URL: The API host used together with endpoint paths

Custom API integration settings and AgentFlow playbook configuration serve different purposes.

Custom API integration controls what the API can do and what fields are available. To control when the agent should collect information, call the API, or use the API response, update the AgentFlow playbook.

Saving an integration does not automatically enable it for every agent. You must enable the integration and its endpoints for each agent under Configuration → Actions, then reference the API in playbook instructions.

Set up a Custom API integration

Use the Custom API builder to create an integration from API information you provide. You can describe the API use case, upload supported API documentation, paste API documentation into the setup chat, or paste a raw cURL command.

SleekFlow uses this information to generate an editable schema. You should always review the generated schema before saving the integration.

Step 1: Start a new Custom API integration

1. Click on the icon on the left navigation bar to go to the SleekFlow AI page
2. In the SleekFlow AI page, click on “Integrations” on the top navigation bar
   
3. You will be redirected to the “Integrations” tab
4. Click “Add integration” at the top right corner
   

This opens the setup flow for connecting an external API. Use the conversational setup window on the left to define your integration.

Instead of filling out a form manually, you can guide AI by providing API details, describing your use case, or pasting an existing API request.

The schema generated by Copilot will appear in the schema canvas on the right. You should always review and refine the generated schema before saving the integration.

For example, you can describe use cases such as:

• Create a payment link for a customer
• Retrieve an order status from your CRM
• Search available booking times
• Update a customer profile in an external system

The description helps Copilot understand the API action you want to set up.

For public API use cases, Copilot may suggest a relevant endpoint for you to confirm before generating the schema. For internal or private APIs, provide API documentation, endpoint details, or a raw cURL command so Copilot can generate the schema based on your own API setup.

Step 2: Add API information

You can provide API information in different ways, depending on what you already have.

Option 1: Describe the API use case

Use this option when you know what you want the agent to do, but do not have the full API schema ready.

Enter a short description of the API action in the setup chat. For example:

“Create a Stripe payment link for a customer based on the selected product and customer email.”

Copilot may use your description to suggest a relevant endpoint or ask follow-up questions before generating the schema.

Option 2: Upload API documentation

Use this option when you have API documentation in a supported file format.

Supported upload formats include:

• .JSON
• .YAML
• .YML
• .PDF

Each uploaded file can be up to 1 MB.

After you upload the file, Copilot reviews the documentation and uses it to generate the schema.

Option 3: Paste API documentation

Use this option when you only need to provide part of your API documentation.

Paste the relevant endpoint details into the setup chat, such as:

• Endpoint URL
• Request method
• Required headers
• Query parameters
• Path parameters
• Request body
• Example response

This is useful when the full API documentation is too long, or when you only want to set up one specific endpoint.

Option 4: Paste a raw cURL command

Use this option when you already have a working API request from API documentation, browser developer tools, Postman, or an engineering handoff.

Paste the raw cURL command into the setup chat. Copilot can use the cURL command to identify request details and generate the schema.

Note: Raw cURL input is an alternative setup method. You should still review the generated schema before saving the integration.

Step 3: Review and refine the schema

After Copilot generates the schema, review the fields in the schema canvas before saving the integration.

The schema defines how AgentFlow understands and uses the API. Clear field descriptions and accurate settings help the agent call the API correctly during customer conversations.

For each endpoint, review the following sections:

Review the Base URL

The Base URL is displayed in the schema canvas after setup.

Review it carefully before saving the integration, especially if your API has different environments, such as staging and production.

The Base URL works together with the endpoint path to determine where the API request is sent.

Warning: If the Base URL is incorrect, API calls may fail or be sent to the wrong environment.

Review inputs and data source

In the Inputs section, check the fields that the API needs to receive.

For each input, review:

• Field name
• Description
• Data source
• Static value, if used

The field description should clearly explain what value is needed. This helps the agent map customer information to the correct API field.

Review outputs and Expose to AI

In the Outputs section, check the fields returned by the API.

Turn on Expose to AI only for fields the agent should use when generating customer replies.

Turn it off for fields that:

• Are not useful for the conversation
• Are only needed for backend processing
• Contain sensitive information
• May confuse the agent if exposed

Tip: Expose only the output fields that are needed for the agent’s response. This helps keep replies focused and reduces unnecessary information.

Step 4: Define how input values are provided

In the Inputs section, use the Data Source column to choose how each input value should be provided.

In the “Inputs” section, use the Data Source column to choose how each input value should be provided. You can choose “Collect from chat” or “Static value”.

Collect from chat

Select “Collect from chat” when the value should come from the customer during the conversation.

Use this for information that may be different for each customer or request, such as email, booking date, product preference, or order number.

When you use “Collect from Chat”:

1. Make sure the field description clearly explains what information is needed.
2. Update the AgentFlow playbook so the agent knows when to ask the customer for this information.
3. Test the agent to confirm the collected value is mapped to the correct API field.

Static value

Select “Static value” when the same value should be sent every time the API is called.

Use this for fixed values such as a default system parameter, region code, internal flag, or channel identifier.

When you use “Static value”:

1. Enter the fixed value in the “Value” field.
2. Check that the value matches the format required by your API.
3. Save your changes before testing the integration in AgentFlow.

Step 5: Save and publish the integration

Before saving, review the integration details to make sure they are accurate.

Check that:

• the endpoint schema is correct
• each input has the right description and data source
• static values are entered correctly, if used
• output descriptions are clear
• only the right output fields are exposed to AI

When you are ready, click “Save”.

After saving, the Custom API integration becomes available for supported SleekFlow AI modules, such as AgentFlow.

Managing your Custom API integration

You can manage your Custom API integration once it’s added to SleekFlow.

Follow the steps below to manage your Custom API integrations:

1. Click on the icon on the left navigation bar to go to the SleekFlow AI page
2. In the SleekFlow AI page, click on “Integrations” on the top navigation bar
3. You will be redirected to the “Custom API integration” page, where you will find the Custom API added to your SleekFlow’s workspace
   

You can open a Custom API integration to review its configuration, including the integration name, authentication settings, Base URL, endpoint schema, inputs, outputs, and assigned agents.

Editing Custom API

After creating a Custom API integration, you can update its configuration at any time. This allows you to refine inputs, adjust outputs, or update authentication as your API or use case evolves.

To edit a Custom API:

1. In the Custom API integration dashboard, locate the Custom API you would like to edit
2. Click on the Custom API
3. You will be redirected to the Custom API details page

Updating the Integration name

You can rename your Custom API integration to better reflect its purpose. Use a clear and descriptive name so it is easy to identify when selecting the integration in AgentFlow.

Reviewing the Base URL

The Base URL is visible on the Custom API details page.

Use this field to check which API host or environment the integration uses. This is especially useful if your API has different environments, such as staging and production.

You no longer need to ask the builder agent to retrieve the Base URL after the integration is created.

Updating authentication settings

In the “Authentication” section, you can configure or update your API credentials.

• Click “Update key” to modify your API key or bearer token
• Make sure the authentication method matches your API requirements

Changes here affect how all API requests are authorized.

Edit endpoint schema

In the “Endpoint schema section”, you can review the endpoints included in the Custom API integration.

Each endpoint shows the request method, endpoint path, and a short description of what the endpoint does. You can expand an endpoint to review its configuration. Review these details together with the Base URL to make sure the request is sent to the correct API endpoint.

Click “Add / Edit endpoints” to update the endpoint schema, add endpoints, or remove endpoints if your API setup changes.

Update playbook command name

Each endpoint has a Playbook command name.

This is the command name used when configuring the integration in AgentFlow. Use a clear, recognizable name that describes the action, such as /get_available_slots or /create_booking.

If you rename a command that is already used by an agent, review the related AgentFlow setup to make sure the agent still references the correct command.

Edit inputs

In the Inputs section, review the fields that the API needs to receive.

For each input, you can:

• update the description
• choose the “Data source”
• enter a value if the data source is set to Static value

Input descriptions should clearly explain what value is needed. This helps the agent map collected information to the correct API field.

Edit outputs

In the Outputs section, review the fields returned by the API.

For each output, you can:

• update the description
• turn Expose to AI on or off

Turn on Expose to AI for fields the agent should be able to use in customer replies. Turn it off for fields that are not useful for the conversation or should not be shown to the agent.

Tip: Expose only the output fields that are needed for the agent’s response. This helps keep replies focused and reduces unnecessary information.

Review assigned agents

At the bottom of the page, check Assigned agents to see which agents are currently using this integration.

If an agent is listed, you can manage how the integration is used from the agent setup.

Save your changes

After making updates, click “Save changes” at the bottom of the page.

Your updates will apply to the Custom API integration. After saving, review and test any AgentFlow setups that use this integration, especially if you changed command names, endpoint details, inputs, outputs, or exposed fields.

Deleting Custom API

To delete a Custom API:

1. In the Custom API integration dashboard, locate the Custom API you would like to delete
2. Click on the icon next to the Custom API
3. Select “Delete” 
   
4. A warning message will appear. Click “Delete” to confirm the action