Zendesk | ElevenLabs Documentation

Overview

Connect your ElevenLabs AI agents with Zendesk to manage support tickets, users, and organizations. This integration enables your agents to create and update tickets, search for existing records, manage users, and respond to incoming ticket comments.

Setup

This integration uses Zendesk API token authentication (email + API token).

Enable API token access

In Zendesk Admin Center, go to Apps and integrations > APIs > Zendesk API and ensure Token Access is enabled.

Generate an API token

Go to Apps and integrations > APIs > API tokens and click Add API token. Copy the token immediately — it will not be shown again after you close the dialog.

Find your subdomain

Your Zendesk subdomain is the first part of your Zendesk URL (e.g., mycompany from mycompany.zendesk.com).

Connect in ElevenLabs

In the ElevenLabs integration setup, enter your email, API token, and subdomain.

Demo video

This demo uses legacy webhook tools. If you’re using the native Zendesk integration, the tools are configured automatically — no manual webhook setup is needed.

Zendesk Integration Demo

How it works

Ticket management

The agent creates, updates, and searches tickets during a conversation — adding comments, changing status, assigning tickets, and managing tags. Bulk operations on up to 100 tickets at a time are supported.

Customer lookup

The agent searches for users by email or phone, retrieves their ticket history, and creates new user records when needed.

Automated responses

The agent monitors and responds to incoming ticket comments, providing first-line support. You can optionally set a daily ticket limit and reference an existing Zendesk trigger for custom routing conditions.

Call logging

For voice-based interactions, the agent logs call activity directly in Zendesk and checks agent availability.

Example system prompt

View example system prompt

You are a helpful support agent responsible for gathering information
from users and creating support tickets. Be friendly, precise, and
concise.
Begin by asking for a detailed description of the problem. Then, ask
relevant support questions to gather additional details, one question
at a time, and wait for the user's response before proceeding.
Once you have a description of the issue, say you will check if there
are similar issues and any known resolutions. Search for resolved
tickets with similar issues and review their comments to find
relevant learnings.
After this, tell the customer the recommended resolution from a
previous similar issue. If they have already tried it or still want
to move forward, proceed to ticket creation. Only provide resolution
advice derived from previous ticket comments.
After capturing the support issue, gather the following contact
details:
- The user's name
- A valid email address (note: the email may be transcribed from
  voice, so ensure it is formatted correctly)
- Read the email back to the caller to confirm accuracy
Once the email is confirmed, explain that you will create the ticket.
Create the ticket with all relevant issue details in the comment body.
Thank the customer and say support will be in touch.
Clarifications:
- Do not inform the user that you are formatting the email; simply
  do it.
- If the caller asks you to move forward with creating the ticket,
  do so with the existing information.
Guardrails:
- Do not speak about topics outside of support issues.

Legacy webhook setup

If you use the native Zendesk integration, tools are configured automatically. The steps below apply only to manual webhook setup.

The integration uses three webhook tools to create the support agent. Review each tool’s configuration in the tabs below.

zendesk_get_ticket_comments

zendesk_get_resolved_tickets

zendesk_open_ticket

Name: zendesk_get_ticket_comments Description: Retrieves the comments of a ticket. Method: GET URL: https://acmecorp.zendesk.com/api/v2/tickets/{ticket_id}/comments.json

Headers:

Content-Type: application/json
Authorization: (Secret: zendesk_key)

Path Parameters:

ticket_id: Extract the value from the id field in the get_resolved_tickets results.

Tool JSON:

1 {
2   "type": "webhook",
3   "name": "zendesk_get_ticket_comments",
4   "description": "Retrieves the comments of a ticket.",
5   "api_schema": {
6     "url": "https://acmecorp.zendesk.com/api/v2/tickets/{ticket_id}/comments.json",
7     "method": "GET",
8     "path_params_schema": [
9       {
10         "id": "ticket_id",
11         "type": "string",
12         "description": "Extract the value from the id field in the get_resolved_tickets results.",
13         "dynamic_variable": "",
14         "constant_value": "",
15         "required": false,
16         "value_type": "llm_prompt"
17       }
18     ],
19     "query_params_schema": [],
20     "request_body_schema": null,
21     "request_headers": [
22       {
23         "type": "secret",
24         "name": "Authorization",
25         "secret_id": "zendesk_api_token"
26       },
27       {
28         "type": "value",
29         "name": "Content-Type",
30         "value": "application/json"
31       }
32     ]
33   },
34   "response_timeout_secs": 20,
35   "dynamic_variables": {
36     "dynamic_variable_placeholders": {}
37   }
38 }

Ensure that you add your workspace’s Zendesk secret to the agent’s secrets.

Evaluation configuration

Configure evaluation criteria and data collection parameters in the Analysis tab to monitor customer interactions.

Evaluation criteria for support interactions — Track how well the AI agent performs against key evaluation criteria like issue relevance, sentiment, and resolution success.

Monitoring results

After setting up the integration, review conversation summaries and evaluation data in the ElevenLabs dashboard to track agent performance.

Support agent conversation summary — Get a high-level overview of each conversation and listen to the conversation's audio recording.

Data collection parameters from conversation transcripts — Monitor the data collected during each interaction, including tools used, issue details, and customer information.

Detailed conversation transcript example — Review detailed transcripts of conversations to understand agent performance, tool usage and customer interactions.

Security considerations

Use HTTPS endpoints for all webhook calls.
Store sensitive values as secrets using the ElevenLabs Secrets Manager.
Validate that all authorization headers follow the required format.