Go to Kayako Classic
Overview
Kayako does not currently provide an official MCP (Model Context Protocol) server or a native Claude connector for direct Kayako-to-Claude integration.
Users exploring AI workflows with Kayako may expect a built-in way to connect Claude directly for tasks such as:
- generating draft replies from incoming tickets,
- grounding AI responses in internal documentation,
- or exposing ticket data for reporting and analysis.
At this time, that direct MCP-based connection is not a supported out-of-the-box capability in Kayako. When this type of workflow is needed, it must be built through Kayako’s existing integration surfaces and any required external middleware.
Solution or Explanation
If the goal is to connect Kayako with Claude, the key point is this:
Kayako supports integrations through its existing tools, but not through an official MCP connector.
What is supported today
The currently documented integration options are:
-
Kayako API
Use the API when you need custom access to conversations, users, organizations, or other data for an external workflow. -
Endpoints
Use Endpoints when Kayako needs to send selected data to an external system through automation. -
Webhooks
Use Webhooks when an external system needs to send event data into Kayako timelines. -
Zapier
Use Zapier for supported no-code or low-code integrations across connected applications.
What this means for Claude integrations
If you want Claude to participate in a support workflow, the Claude-specific parts must be handled outside Kayako. This includes items such as:
- the MCP layer,
- prompt design,
- retrieval of internal documentation,
- business rules for agent review,
- and any reporting or analytics logic exposed to managers.
In practice, the implementation usually looks like this:
- Retrieve or send Kayako data using the API, Endpoints, Webhooks, or Zapier, depending on the workflow.
- Pass that data to your own middleware or integration layer.
- Let that external layer handle:
- communication with Claude,
- retrieval of your internal knowledge sources,
- response generation,
- and any approval or review workflow before an agent sends a reply.
Kayako can act as the data source or part of the automation flow, but it does not currently provide the direct MCP connector itself.
Example
A common use case is having an AI assistant draft suggested replies for agents.
In that setup, Kayako can provide the ticket data through the API or automation-driven delivery, but the following pieces still need to be built and managed outside Kayako:
- retrieving internal documentation,
- sending the combined context to Claude,
- applying review rules,
- and returning the generated draft to the agent workflow.
The same principle applies to manager-facing reporting assistants: Kayako can provide the source data, but the Claude-facing query experience must be implemented externally.
Verification
You can verify whether this article applies by checking your planned design:
- If your workflow depends on a direct built-in MCP endpoint or native Claude connector inside Kayako, that capability is not currently available.
- If your workflow can use Kayako’s API, Endpoints, Webhooks, or Zapier together with your own middleware, then the integration remains possible as a custom implementation.
Additional Notes
- The Kayako AI Agent Plugin is a separate capability and should not be treated as an MCP connector for Claude.
- The AI Agent Plugin is not generally available and is currently described as being in a limited Proof of Concept phase.
- A feature request for official MCP support has been submitted to the product team for review.
- There is no ETA or guarantee for feature-request implementation.
- If official MCP support is important for your organization, submit a feature request through support so that demand for the capability is recorded.
Syed Haider
Comments