The Zendesk MCP Server acts as a smart bridge that lets AI models interact directly with a company's Zendesk support platform. In simple terms, it allows an AI assistant to read, organize, and respond to customer support tickets just like a human agent would. By connecting these systems, teams can automate the routine parts of customer service, ensuring that every customer gets a fast, context-aware response without a developer having to manually copy and paste data between windows. On a more technical level, this server provides a robust toolkit for comprehensive ticket management. It enables the retrieval of ticket lists with full pagination and sorting, the ability to view or post internal and public comments, and the functionality to update ticket fields like priority, status, and assignees. Beyond basic data operations, it features specialized prompts designed to help the AI perform deep-dive ticket analysis and draft professional, nuanced responses based on the specific history of a customer's interaction. What makes this tool particularly valuable for developers is its ability to turn the Zendesk Help Center into a live knowledge base for an LLM. By exposing help articles through a dedicated resource URI, the server allows an AI to pull from official documentation to ensure its answers are grounded in the company's actual policies and guides. Built for modern workflows, the server supports both local execution via Python and containerized deployment with Docker, making it easy to embed real-time customer support intelligence into any AI-driven application or desktop workflow.
Category: CRM, ERP & E-commerce
Tags: Customer Support, helpdesk, knowledge-base, ticketing, zendesk
uv or via Docker. Using uv: 1. Navigate to the project directory. 2. Build the environment: uv venv && uv pip install -e . (or use the shortcut uv build). 3. Set up your Zendesk credentials in a .env file (refer to .env.example for the required format). Using Docker: 1. Copy .env.example to .env and fill in your Zendesk credentials. 2. Build the image: bash docker build -t zendesk-mcp-server . 3. Run the server: bash docker run --rm --env-file /path/to/.env zendesk-mcp-server Note: Use -i when connecting to MCP clients over STDIN/STDOUT. ---json { "mcpServers": { "zendesk": { "command": "uv", "args": [ "--directory", "/path/to/zendesk-mcp-server", "run", "zendesk" ] } } }settings.json: json { "mcpServers": { "zendesk": { "command": "/usr/local/bin/docker", "args": [ "run", "--rm", "-i", "--env-file", "/path/to/zendesk-mcp-server/.env", "zendesk-mcp-server" ] } } } ---page (int), per_page (int), sort_by (string), sort_order (string). * get_ticket: Retrieve a specific Zendesk ticket by its ID. * Parameters: ticket_id (int). * get_ticket_comments: Retrieve all comments for a specific Zendesk ticket. * Parameters: ticket_id (int). * create_ticket_comment: Create a new comment (public or private) on an existing ticket. * Parameters: ticket_id (int), comment (string), public (boolean). * create_ticket: Create a new Zendesk ticket. * Parameters: subject (string), description (string), requester_id (int), assignee_id (int), priority (string), type (string), tags (array), custom_fields (array). * update_ticket: Update fields on an existing ticket (status, priority, assignee, etc.). * Parameters: ticket_id (int), subject (string), status (string), priority (string), type (string), assignee_id (int), requester_id (int), tags (array), custom_fields (array), due_at (string). ---draft-ticket-response prompt. Example: An agent asks the AI, "Help me respond to Ticket #1024 about a billing error." The AI uses get_ticket to understand the issue, searches `zendesk:/knowledge-basefor refund policies, and presents a complete draft for the agent to review and send viacreate_ticket_comment`.
get_tickets and analyze-ticket tools, a user can instruct the AI to scan the latest 50 "New" tickets. The AI can identify the sentiment, summarize the core issue, and automatically update ticket fields (like priority or tags) to ensure the most frustrated or critical customers are handled first. Example: A manager tells the AI to "Scan all new tickets from the last hour. If any mention 'system down' or 'data loss,' set their priority to 'Urgent' and tag them as 'Critical_Escalation'." The AI iterates through the list and uses update_ticket to modify the necessary records.get_ticket_comments and compare the recurring questions against the existing `zendesk:/knowledge-base`. It can then identify topics that are frequently asked about but not well-documented. Example: A documentation lead asks, "Analyze the last 100 resolved 'Question' tickets. Are there any common issues that aren't covered in our Help Center articles?" The AI identifies a pattern of users struggling with "API Key rotation" and suggests a new article title and outline.
Part of MCP Servers
You can install it by cloning the repository and running uv venv and uv pip install -e . or by building the provided Dockerfile. You must then register a public OAuth client in your Zendesk Admin Center, set your subdomain and client ID in a .env file, and execute uv run zendesk-auth once to authorize your local environment.
The server enables LLM clients to fetch ticket lists, inspect individual tickets, retrieve comment threads, create tickets, and post public or private comments. It also exposes prompts for ticket analysis and drafting responses, and serves the entire Zendesk Help Center documentation as a knowledge base resource.
It works with any Model Context Protocol client that supports local command-line or Docker-based server connections over standard I/O, including Claude Desktop and Claude Code.
The server uses per-operator public OAuth with PKCE by default, ensuring API actions reflect the specific user permissions and author identity in Zendesk. Access tokens are stored locally and automatically refreshed before expiry or upon invalid token errors. Legacy Zendesk API tokens are also supported.
Yes, Zendesk MCP Server is open source and distributed under the Apache 2.0 license.