Zendesk MCP Server (fruggr)
Zendesk support tickets and help center articles managed via natural language queries.
Data last scanned today · Reviewed today
Overview
The Zendesk MCP Server (fruggr) allows AI assistants to interact directly with Zendesk workflows. Users can search and retrieve Help Center articles, manage ticket lifecycles—including triage, status updates, and comment thread management—and process image attachments from tickets directly within their AI interface. The server operates using individual user permissions rather than a shared administrative key, ensuring that actions remain within the scope of the authenticated user's access. It supports complex tasks like maintaining knowledge-base articles by enabling section-by-section revisions to avoid large HTML round-trips.
Our verdict
With a security score of 68/100, this server is classified as medium risk. It is a community-maintained project rather than an official Zendesk release, and it has not yet undergone independent editorial review. While it implements secure OAuth authentication and offers a read-only configuration option, it is best suited for users who are comfortable with third-party integrations and who plan to restrict access permissions appropriately.
- Supports a read-only configuration to minimize accidental modifications.
- Uses OAuth for secure authentication instead of shared keys.
- Small dependency footprint with only 15 direct packages.
- Maintains an active development cycle with recent commits.
- Community-maintained project lacking official vendor support.
- Limited community adoption with only 3 GitHub stars.
Setup
{
"mcpServers": {
"fruggr-zendesk-mcp-server": {
"args": [
"-y",
"@fruggr/zendesk-mcp-server"
],
"command": "npx"
}
}
}Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.
Tools
| Tool | Description | Risk |
|---|---|---|
| get_article | Reads a specific article by ID. | low |
| list_promoted_articles | Lists promoted (featured) articles. | low |
| get_article_outline | Returns the structural outline of an article. | low |
| get_article_section | Reads a specific section of an article. | low |
| update_article_section | Updates a specific section of an article. | medium |
| get_ticket | Retrieve a ticket by ID with optional comments and its live SLA state (resolved via a scoped search) | low |
| get_ticket_history | Read a ticket's change history (audit trail) as a chronological, oldest-first timeline of who changed what and when — field changes as before → after with actor names, comment presence (not bodies), system noise filtered; cursor-paginated | low |
| get_ticket_attachments | Download ticket attachments — images delivered as native multimodal content for the client's own model to analyze; oversize/over-limit images and non-images come back as text references | low |
| search_tickets | Search tickets using Zendesk query syntax, with per-result SLA state | low |
| list_tickets | List tickets with cursor-based pagination | low |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›How does this server handle user authentication?
This server uses OAuth 2.1 PKCE, requiring each user to authenticate with their own Zendesk credentials to ensure the assistant only accesses data the user is permitted to see. Static API tokens are not supported.
›Can I prevent the assistant from modifying Zendesk data?
Yes, the server supports a read-only mode, which restricts all operations to read-only access and prevents any modifications to your Zendesk data.
›How are large Help Center articles handled?
The server supports section-based editing, which allows you to read and rewrite articles one section at a time based on headings rather than processing the entire HTML body.
›Do I need a separate API key for image analysis in tickets?
No, this server delivers ticket images as native MCP multimodal content directly to the client's own model, requiring no server-side vision model or additional API keys.
›What are the deployment options for this server?
You can run the server locally on your machine via stdio for tools like Claude Desktop and VS Code, or deploy it as a private remote MCP server over HTTP.
›Can I limit which tools the assistant has access to?
Yes, tools are segmented into namespaces that you can selectively enable, allowing you to load only the specific tool surfaces required for your context.
Badge
Maintain this server? Add the live badge to your README.