Introduction

The MCP (Model Context Protocol) component is a native tool integrated into wolkvox Studio that allows connecting routing point flows with external services such as CRM, calendars, ERPs, or other platforms using the standard MCP protocol.

This component functions as a smart bridge between wolkvox and external systems, enabling conversational flows to query information, create records, or execute actions in other systems without the need to develop custom code.

Among its main benefits are:

• Low-Code Integration: Allows automatically importing available functions from external servers, reducing development time.
• Native Omnichannel: Can be used in routing points of type IVR, Chat, Interactions, and Agent Scripting.
• Self-Service Automation: Facilitates the creation of conversational flows that allow, for example, querying information or scheduling appointments in real time.

To use the MCP component, you must first configure the connection with the external MCP server from wolkvox Manager and then use the component within wolkvox Studio.

Configuration

Create the MCP Connection from wolkvox Manager

Before using the MCP component in a flow, you must first configure the connection with the external MCP server.

1. Click on the configuration icon located in the top right corner of wolkvox Manager.
2. In the side menu, select Integrations.
3. Then open the MCP Client tab.
4. Click on New Connection to create the integration with the MCP server.
5. You must complete the following fields:
   1. Connection Name: Identifier name for the integration.
   2. MCP Endpoint URL: URL of the external MCP server.
6. Authentication Type: Can be:
   • none
   • bearer
   • oauth

Depending on the selected method, additional fields will be enabled.

1. If you choose Bearer:
   • The Bearer Token field will appear, where you must enter the authentication token.
2. If you choose OAuth:
   • You must click on Connect OAuth Account to start the authorization process with the external service.
   • During this process, an authentication window from the external service will open where access must be approved.
   • Once completed, the connection status should be displayed as:
   • oauth_ok
   • This indicates that the connection was successful.

Import the Available Tools from the MCP Server

Once the connection is created, you must import the functions available on the MCP server.

1. In the MCP Tools section, click on: List Tools
   • This will query the MCP server and display the catalog of available functions.
2. For example, in integrations with platforms like Notion or Zapier, tools such as: notion-search,notion-fetch, notion-create-pages, notion-update-page, notion-create-database, notion-create-comment, notion-get-users
   • Mark the tools you want to enable within wolkvox.
   • These tools will be the ones that can be executed from the wolkvox Studio flows.

Click on Save Configuration to store the selected tools.

Use the MCP Component within wolkvox Studio

Once the integration is configured, you can use the component within your conversational flows. Access wolkvox Studio and open the routing point where you want to use the integration.

1. The MCP component is found within the component group: Cognitive
2. Drag the MCP Client component to the flow canvas in the position where you want to execute the integration.

Configure the MCP Component in the Flow

To configure the component:

1. Double-click on the MCP component in the flow.
2. The configuration window will open. The available fields are:
   • MCP Connection: Allows you to select the connection previously configured in wolkvox Manager.
   • Base Instruction: This is the instruction that will be sent to the MCP server.
     • Here you can use flow variables.
     • For example, in routing points of type chat, you can use the variable: $txt_query
       • This variable corresponds to the message that the customer writes in the chat.
     • The MCP server will interpret the instruction and execute the corresponding action in the external system.

Test the Integration in the Chatbot

Once the flow is configured, you should test the integration using the Test ChatBot tool in wolkvox Studio.

In the example shown in the tests:

1. An MCP connection with Zapier was used, which integrated Google Calendar.
2. The user wrote in the chat: What can you do for me?

The chatbot responded by indicating the available actions, such as:

• Create events in Google Calendar
• Search for existing events
• Create spreadsheets
• Add rows to spreadsheets

Later, a request was made to create an event with the following message:

The system processed the request and automatically created the event in Google Calendar.

Possible Time Differences in Created Events

During testing, it was observed that the event was created one hour later than the indicated time. This behavior is usually due to issues related to time zones (timezone) in the integration. The most common causes are:

Timezone Difference Between Systems

Google Calendar uses the time zone configured in the account or calendar, while the LLM or MCP server may be using UTC or a different time zone.

Automatic Time Conversion

Some platforms automatically convert received times to UTC before sending them to Google Calendar.

This can cause discrepancies of 1 hour, especially in regions with daylight saving time or different regional settings.

How to Correct Time Differences in the Created Event

In these cases, the client can correct the discrepancy directly from the conversation with the assistant. If the event was created at an incorrect time, the user can request the adjustment by explicitly indicating the required change. For example, if the meeting was scheduled one hour later than requested, the client can write a message like: "The meeting time was set one hour later than requested, please set it one hour earlier." The system will interpret the instruction and execute the corresponding modification in the calendar through the MCP integration, allowing these types of inconsistencies to be quickly corrected without the need for manual intervention in the external system.

Handle Errors Using the Response Variable

The MCP component automatically saves the operation result in the variable: $mcp_client_result

Within this variable is the response generated by the MCP server. For example: $mcp_client_result["response"]

This value can be used within the flow to:

• Validate if the operation was successful.
• Detect integration errors.
• Execute alternative routes.

You can save this result in an additional variable and use different wolkvox Studio components such as Intents to handle different scenarios based on the received response.

Important Considerations

Before using the MCP component, you should consider:

• External Connectivity: The target MCP server (e.g., Salesforce, Zapier, or Google) must be correctly configured and accessible from wolkvox.
• Variable Configuration: It is important to correctly map the input variables of the flow to the MCP component to ensure that instructions are interpreted correctly.