You can use a Zendesk MCP server to ask Claude or ChatGPT about your support performance, summarise ticket conversations or update tickets. Which server you choose depends on what you want your assistant to do.
For support reporting and performance analysis, Geckoboard’s Zendesk Metrics MCP gives your assistant calculated metrics such as first reply time, CSAT, backlog and SLA attainment. For reading conversations and taking actions on tickets, Swifteq and community-built servers provide access to Zendesk records and tools.
Zendesk also has an MCP client. This lets workflows inside Zendesk use tools from external systems, rather than giving an external assistant access to Zendesk.
Here’s how the options compare.
| What you want to do | Which option fits | How it works |
|---|---|---|
| Analyse support performance and prepare regular reports | Geckoboard’s Zendesk Metrics MCP | Gives your assistant calculated support metrics using the same metrics engine as Geckoboard dashboards |
| Read ticket conversations, investigate customer issues and perform ticket actions | Swifteq’s MCP Server for Zendesk | Connects your assistant to Zendesk tickets, Help Centre content and other tools |
| Build an integration around your own requirements | Community MCP servers | Let your developers host and maintain a server with the tools you choose |
| Let Zendesk workflows interact with other systems | Zendesk’s MCP client | Makes external MCP tools available inside Zendesk action flows |
What is a Zendesk MCP server?
MCP stands for Model Context Protocol. It gives AI assistants a standard way to request information and use tools provided by another service.
A Zendesk MCP server gives an assistant such as Claude or ChatGPT access to Zendesk data. The server determines what the assistant can retrieve or change. Some servers provide ticket conversations and actions; Geckoboard provides calculated support metrics.
For reporting, look at how the server provides the numbers. Some servers give the assistant ticket records to work with. Geckoboard gives it metrics that have already been calculated.
Choosing a Zendesk MCP for reporting
“Summarise this ticket” and “Why did our first reply time increase?” both sound like questions about Zendesk. Answering them requires different information.
To summarise a ticket, the assistant needs the conversation. It can read the messages, identify the customer’s problem and suggest what should happen next.
To explain a change in performance, it needs reliable figures to compare. If you ask for median first reply time by support group, the result depends on the reporting period, which tickets are included and how first reply time is calculated.
A server that retrieves ticket records can help with counts, feedback analysis and reports based on those records. But retrieving records doesn’t automatically give the assistant a complete dataset or the calculation rules for your support KPIs.
If the assistant has to calculate a metric itself, two problems can follow:
- Incomplete data. It calculates an answer from the records it retrieved without accounting for tickets it hasn’t seen.
- Inconsistent calculations. It makes its own assumptions about metric definitions, exclusions or how records should be grouped.
Either can produce a precise-looking answer that misrepresents your team’s performance.
Geckoboard calculates the metrics before returning them to the assistant. The assistant can then compare periods, investigate changes and explain the figures.
Geckoboard’s Zendesk Metrics MCP
Best for: support reporting, performance analysis and tools that need consistent Zendesk metrics.
Geckoboard’s Zendesk Metrics MCP connects Claude, ChatGPT and other MCP-compatible assistants to calculated Zendesk metrics.
It uses the same metrics engine that powers Geckoboard dashboards. Your AI-generated reports can use the same definitions as your team’s dashboards, without asking the assistant to reconstruct the calculations each time.
For example, you could ask:
Prepare a weekly support review covering tickets created and solved, backlog, first reply time, SLA attainment and CSAT. Compare the results with the previous week and highlight the largest changes.
Then follow up:
Break down the increase in first reply time by support group and channel.
The assistant requests the relevant metrics from Geckoboard and uses the results to build its answer. It doesn’t need to retrieve thousands of ticket conversations to calculate them.
What you can use it for
Investigate performance changes. Ask why a KPI moved, compare it with an earlier period and break down the figures to see where the change occurred.
Prepare regular reports. Use a repeatable prompt or skill to prepare the figures and commentary for your weekly support review.
Analyse data from several tools. Bring Zendesk metrics together with supported sources such as HubSpot, Aircall and Intercom in the same analysis.
Build custom tools. Use Geckoboard’s metrics in internal dashboards, reports and AI agents without rebuilding the reporting logic.
Geckoboard’s MCP is read-only. For Zendesk, it provides metrics without returning ticket conversations or changing tickets. You can connect a ticket-focused MCP alongside it if you also want your assistant to read conversations or draft replies.
How to get started
- Create a Geckoboard account and connect Zendesk.
- Add Geckoboard to your AI assistant. In Claude, use the Connectors directory; in ChatGPT, you can find it in the Plugins directory.
- Ask a question about your support performance.
The Metrics MCP is included with every Geckoboard plan and is available during the free trial. Follow the setup instructions for your assistant.
Swifteq’s MCP Server for Zendesk
Best for: working with ticket conversations, analysing customer feedback and taking actions in Zendesk.
Swifteq’s MCP Server for Zendesk connects Zendesk to assistants including Claude and ChatGPT. It’s a managed service listed on the Zendesk Marketplace, so you don’t need to host your own server.
It supports tasks such as:
- Reading and summarising ticket conversations.
- Investigating contact reasons and customer feedback.
- Creating tickets and adding notes or replies.
- Working with Help Centre articles and macros.
Swifteq also promotes weekly KPI reports, trend analysis and QA reviews. Its access to conversations is useful when you want a report to explain what customers are saying, as well as what’s happening to ticket volumes.
If you’re choosing it for performance reporting, check how it produces the figures and whether they match your team’s metric definitions. A report that summarises customer feedback has different requirements from one that calculates SLA attainment across your support operation.
Swifteq uses OAuth and says it fetches ticket data on demand without retaining a copy. It offers a free allowance of 500 MCP calls per month per Zendesk account. Paid plans provide unlimited calls, starting at €49 per month for the account and first user, plus €9 per additional user.
Community-built Zendesk MCP servers
Best for: teams with developers who want to control the integration themselves.
Open-source projects such as reminia/zendesk-mcp-server provide another way to connect an AI assistant to Zendesk.
You can inspect the code, choose which tools the assistant can use and adapt the server to your workflows.
You’re also responsible for hosting, authentication, permissions and maintenance. Community servers vary in the Zendesk features they support and the actions they allow, so check the documentation for the project you’re considering.
If you’re starting a new implementation, pay particular attention to authentication. Zendesk is retiring Support API tokens:
- Accounts created from 28 July 2026 cannot create or use them.
- From 27 October 2026, existing accounts cannot create new tokens.
- On 30 April 2027, all remaining tokens will stop working.
Choose a project that supports OAuth or explains how it will migrate before the deadline.
Does Zendesk have its own MCP server?
Zendesk has announced a general-purpose MCP server. Its MCP client is already available, but the client and server are separate products.
At Relate in May 2026, Zendesk said its server would connect tickets, knowledge and other data to external AI systems, with early access planned for the summer.
As of 20 October 2026, we haven’t found public setup documentation confirming availability of that general-purpose server. Check Zendesk’s announcement and current early-access programmes for updates.
Two other Zendesk initiatives also use MCP, but serve different purposes.
Zendesk’s MCP client
The client lets Zendesk use tools provided by external MCP servers. For example, a Zendesk action flow could retrieve information from another system or perform an action there.
Zendesk released the MCP client generally on 10 August 2026. The client doesn’t provide the connection you need to ask Claude or ChatGPT about your Zendesk account.
ChatGPT as a customer support channel
Zendesk also has an early-access programme that uses MCP to let businesses provide customer support inside ChatGPT.
This lets customers seek help through ChatGPT. It serves a different audience from an MCP server used by your team to work with tickets or analyse support performance.
Which Zendesk MCP should you choose?
Choose Geckoboard to investigate support performance and prepare reports using consistent, calculated metrics.
Choose Swifteq to work with ticket conversations, analyse customer feedback and perform ticket actions.
Choose a community server if your developers want to build and maintain a custom integration.
Use Zendesk’s MCP client to build workflows inside Zendesk that need tools from other systems.
Want to ask Claude or ChatGPT about your support performance? Connect Geckoboard’s Zendesk Metrics MCP and start with the metrics your team uses to run support.
