Menu
Mattering Automations connector

A Zendesk MCP where the one destructive step ships with its own undo

Connect an account subdomain, an agent email, and an API token, and 45 Zendesk nodes become workflow steps for visual automations and AI agents: tickets with public replies and internal notes, unified search with an uncapped export, users, organizations, macros, views, and incremental exports. The single destructive node, ticket delete, is a soft delete with a 30-day window, and the connector ships the list and restore nodes to undo it.

45 ready-made nodes One credential to connect Works with Claude, ChatGPT, and Matter

What the Zendesk MCP connector does

The Zendesk connector for Mattering Automations is 45 nodes across the Support API: the full ticket lifecycle (list, get, create, update, soft-delete, and restore, with the conversation from a dedicated comments node), unified search with a count node and an uncapped cursor export, users (create, update, suspend, search), organizations, groups, macros with a preview-only apply, views you can execute as saved filters, read-only trigger and automation listings, tag append and remove, satisfaction ratings, ticket field schema discovery, attachment metadata, job status polling, and the incremental exports that mirror tickets, users, and organizations into other systems.

Every node speaks MCP, the open protocol for connecting AI assistants to tools, so Claude, ChatGPT, and any MCP-capable assistant can use them, alongside Matter, the AI assistant built into the platform. An assistant can answer how many urgent tickets are open, pull a ticket's full conversation, draft an internal note for an agent to review, or look up which organization a requester belongs to, without anyone opening the Zendesk agent workspace.

Because Mattering is a work platform rather than middleware, support work lands somewhere useful by default: a scheduled ticket sweep can open tasks in Mattering Productivity for tickets that need attention, incremental exports can keep a Mattering database mirrored with the help desk for reporting, and an AI step can draft internal notes that wait on a review task instead of going straight to a customer.

The safety posture is deliberate. Exactly one node is destructive: Zendesk: Delete Ticket, and even that is a soft delete with a 30-day recovery window, paired with Zendesk: List Deleted Tickets and Zendesk: Restore Deleted Ticket so the whole recovery loop lives on the canvas; permanent deletion is not offered. User deletion is not offered either, because Zendesk documents it as not recoverable: Zendesk: Suspend or Unsuspend User is the reversible alternative. Macro applies are previews that change nothing until you commit them through a ticket update, and business rules (triggers and automations) are read-only.

One connection, three surfaces

  • Visual automations. Drag Zendesk nodes onto the canvas and wire them to triggers, conditions, and other apps.
  • AI agents, yours or ours. These nodes speak MCP, the open protocol, so Claude, ChatGPT, and any MCP-capable assistant can use them, and Matter, the agent built into the platform, uses them natively.
  • The whole platform. Combine with native tasks, projects, and other connectors, no middleware in between.

Every Zendesk node, ready to drop on the canvas

45 nodes, generated from the connector itself so this list always matches what ships. Expand a category to browse.

Read and query nodes 34

Zendesk: Count Search Results

Return only the number of records matching a search query, without fetching them. Cheap way to branch an automation on how many open tickets match a filter.

Zendesk: Count Tickets in View

Return the (possibly cached, estimate-grade) ticket count for one view without fetching tickets. Rate cap: 5 requests per minute per view per agent.

Zendesk: Execute View

Run a view and get its column titles plus a row per ticket (with a partial ticket object each). Rate cap: 5 requests per minute per view per agent, and results may be served from a cache up to 10 minutes old; for high-frequency polling Zendesk points at Incremental Ticket Export instead. For full ticket objects use List Tickets from View.

Zendesk: Export Search Results

Search WITHOUT the 1,000-result cap, cursor paginated, for one record type at a time. Cursors expire after one hour; rate cap 100 requests per minute. Keep page sizes modest (100 is the recommended value; large pages of archived tickets can time out).

Zendesk: Get Attachment Metadata

Fetch one attachment's metadata: file name, content type, size, and a download URL (content_url). Attachment ids appear on comments from List Ticket Comments. Uploading attachments is a multipart binary POST the canvas cannot produce, so it is deliberately not offered.

Zendesk: Get Current User

Return the user the API token belongs to. Needs no configuration, which makes it the connector's credential health check. Caution: Zendesk allows this endpoint for anonymous callers, so a wrong credential can come back as HTTP 200 with a null user id instead of a 401; a healthy connection is status 200 AND a non-null data.user.id. The response never echoes the API token (the authenticity_token field is a CSRF token for help-center calls, not your credential).

Zendesk: Get Group

Fetch one group by id.

Zendesk: Get Job Status

Poll a background job (status: queued, working, completed, failed, or killed, plus per-record results). Honest note: every write in this connector is synchronous, so no node here produces a job; Zendesk returns job statuses from its bulk endpoints (create/update/delete many, merge, permanent delete), which are not built. Use this to poll a job id obtained elsewhere. Job data expires one day after the job is created.

Zendesk: Get Macro

Fetch one macro by id, including its actions list.

Zendesk: Get Organization

Fetch one organization by id.

Zendesk: Get Ticket

Fetch one ticket by id. The response holds the ticket's properties and its first comment as description; the full conversation comes from List Ticket Comments.

Zendesk: Get User

Fetch one user by id.

Zendesk: Incremental Organization Export

Time-based incremental export of organizations created or changed since a Unix epoch start_time (organizations have no cursor variant). Follow next_page / use the returned end_time as the next start_time; a page shorter than 1,000 records means you are caught up. Rate cap: 10 requests per minute.

Zendesk: Incremental Ticket Export

Cursor-based incremental export: every ticket created or changed since a point in time, the right tool for keeping an external system in sync. First call: start_time (Unix epoch, at least one minute in the past). Every later call: the previous response's after_cursor. Stop when end_of_stream is true and reuse the last cursor next run. Rate cap: 10 requests per minute.

Zendesk: Incremental User Export

Cursor-based incremental export of users created or changed since a point in time. Same cursor protocol as the ticket export (start_time first, then after_cursor; stop at end_of_stream). Rate cap: 20 requests per minute (60 with the High Volume add-on).

Zendesk: List Automations

List Zendesk automations (time-based business rules, evaluated hourly) with their conditions and actions. Read-only by design.

Zendesk: List Deleted Tickets

List soft-deleted tickets from the last 30 days that have not yet been scrubbed, oldest first (up to 100 per page). Pair with Restore Deleted Ticket to undo a deletion.

Zendesk: List Group Memberships

List which agents belong to a group (membership records carry user_id, group_id, and whether it is the agent's default group).

Zendesk: List Groups

List agent groups (up to 100 per page). Groups are the routing unit: tickets are assigned to a group, or to an agent within one.

Zendesk: List Macros

List the shared and personal macros available to the authenticated agent (up to 100 per page).

Zendesk: List Organizations

List organizations, cursor paginated (up to 100 per page).

Zendesk: List Satisfaction Ratings

List CSAT ratings received on tickets, filterable by score and time window. Note: these endpoints serve Zendesk's LEGACY CSAT feature; accounts that switched to the newer survey-based CSAT will see no new ratings here.

Zendesk: List Tags

List the account's most-used tags with usage counts (the most popular tags of the last 60 days, most used first; up to 100 per page).

Zendesk: List Ticket Comments

List a ticket's full conversation, oldest first by default (up to 100 per page). Read-only by API design: new comments are added with the Update Ticket node's comment fields, not here.

Zendesk: List Ticket Fields

List every system and custom ticket field with types and dropdown options. This is the schema-discovery node: the field ids and option values here are what the create/update ticket custom_fields arrays refer to.

Zendesk: List Tickets

List tickets in the account, cursor paginated (up to 100 per page). Page through with meta.after_cursor. For filtered sets prefer Search or a view node; for a full one-way sync prefer Incremental Ticket Export.

Zendesk: List Tickets from View

List the FULL ticket objects currently matching a view (up to 100 per page). The heavier sibling of Execute View.

Zendesk: List Triggers

List ticket triggers (event-fired business rules) with their conditions and actions. Read-only by design: this connector does not modify admin configuration.

Zendesk: List Users

List users in the account, cursor paginated (up to 100 per page), optionally filtered by role or external_id.

Zendesk: List Views

List the shared and personal views available to the authenticated agent. A view is a saved ticket filter; execute one to get its current tickets.

Zendesk: Preview Macro on Ticket

Show the full ticket as it WOULD look after applying a macro. Doc wording: "It doesn't actually change the ticket." To commit, pipe the returned result.ticket fields (and result.ticket.comment if present) into an Update Ticket node.

Zendesk: Search

Unified search across tickets, users, and organizations with the Zendesk query language, for example type:ticket status:open tags:vip created>2026-07-01. HARD CAP: at most 1,000 results per query (100 per page; page 11 returns 422), though count reports the true total. For result sets beyond the cap, use Export Search Results. New records can take a few minutes to be indexed.

Zendesk: Search Organizations

Find organizations by exact name (case insensitive) or exact external_id. Provide exactly one of the two; the API rejects the request otherwise. For fuzzy matching use the Search node with type:organization.

Zendesk: Search Users

Search users by name, email, phone, notes, or details, for example the full search syntax jane@example.com or role:agent smith. Offset pagination only, capped at 10,000 records per query (100 per page). external_id lookups take the id verbatim, not search syntax.

Action nodes 11

Zendesk: Add Ticket Tags

APPEND tags to a ticket without touching its existing tags (HTTP PUT on the tags subresource; the Update Ticket tags field replaces the whole set instead). Returns the ticket's full updated tag list. Does not work on closed tickets.

Zendesk: Create Organization

Create an organization. The name must be unique in the account (a duplicate returns 422). Users whose email domain matches a domain_names entry can be auto-associated.

Zendesk: Create Ticket

Create a ticket. The comment body becomes the ticket's description (first comment). Returns 201 with the new ticket. Business rules (triggers) run on creation and can email the requester. Reversible: a created ticket can be soft-deleted and restored within 30 days.

Zendesk: Create User

Create a user (defaults to an end user). Set skip_verify_email true to avoid sending the verification email, which most automations want. Creating agents or admins can add billable seats. Reversible in workflow terms via suspension; note this connector deliberately ships no user delete (Zendesk user deletion is not recoverable).

Zendesk: Delete Ticket

DESTRUCTIVE: soft-delete a ticket. The ticket moves to the deleted-tickets list, stays recoverable for up to 30 days (List Deleted Tickets + Restore Deleted Ticket nodes), and is then permanently scrubbed. Rate cap: 400 deletions per minute. Permanent deletion is deliberately not offered by this connector.

Zendesk: Remove Ticket Tags

Remove the listed tags from a ticket, leaving other tags in place. Returns the remaining tag list. Reversible (add them back); does not work on closed tickets.

Zendesk: Restore Deleted Ticket

Restore a soft-deleted ticket back to the account. Works only within the 30-day recovery window and only for tickets still on the deleted-tickets list.

Zendesk: Suspend or Unsuspend User

Set a user's suspended flag: true blocks sign-in and routes their new tickets to the suspended queue, false lifts the suspension. Fully reversible; this is the safe alternative to user deletion (which Zendesk documents as not recoverable, and which this connector does not offer).

Zendesk: Update Organization

Update an organization. Only filled fields are sent; tags and domain_names replace their whole lists when provided.

Zendesk: Update Ticket

Update a ticket: add a public reply or internal note (comment_body + comment_public), change status, reassign (assignee_id / group_id), set priority, or replace tags. Only the fields you fill are sent. This is THE way to add a comment in the Zendesk API. Rate cap: 30 updates per 10 minutes per ticket. Closed tickets cannot be updated.

Zendesk: Update User

Update a user's profile. Only filled fields are sent. Zendesk quirk: passing email here ADDS it as a secondary, unverified identity rather than replacing the primary address. Rate cap: 5 updates per minute per user.

Connecting Zendesk takes one credential

Zendesk issues API tokens in Admin Center under Apps and integrations > APIs > Zendesk API, where token access must also be enabled before a token works. A token is paired with an agent's email address and acts as that agent, so a dedicated integration agent is the sensible identity. The Mattering credential packs three parts into two fields: the username is your account subdomain and the agent email joined with a pipe, for example examplecorp|jane@example.com, and the password is the API token itself.

1

In Zendesk Admin Center, open Apps and integrations > APIs > Zendesk API, enable token access in Settings, and add an API token. Copy it when it is shown.

2

In Mattering Automations, open Connectors, click Add Connector, choose Zendesk, and create a credential: username is subdomain|email (for example examplecorp|jane@example.com), password is the API token.

3

Click Test Connection. Mattering calls the current-user endpoint and requires a real user identity in the response, because that endpoint answers anonymous callers with a 200 instead of an error when the credential is wrong.

4

All 45 Zendesk nodes appear in the canvas palette. Run Zendesk: Get Current User first: it takes no parameters and confirms which agent your automations act as.

Common uses for the Zendesk MCP

Triage sweeps that end as tasks

Zendesk: Search finds tickets with the full query language, for example type:ticket status:open priority:urgent, and Zendesk: Count Search Results branches an automation on the number without fetching anything. Views work too: Zendesk: List Views shows the saved filters agents already use, and Zendesk: List Tickets from View returns their full ticket objects. Route what matches into Mattering Productivity tasks so nothing sits unseen.

Replies and internal notes, drafted safely

The Zendesk API has no separate comment-create endpoint: comments are added through ticket update, and Zendesk: Update Ticket carries typed fields for the comment body and its public flag. A public reply emails the requester; an internal note stays inside the team, which makes it the safe path for anything AI-drafted. Zendesk: List Ticket Comments reads the full conversation back, oldest first.

Search past the 1,000-result ceiling

Zendesk: Search caps at 1,000 results per query by API design, even though its count reports the true total. Zendesk: Export Search Results is the uncapped path: cursor-paginated, one record type at a time, with cursors that expire after an hour. Between the two, there is no result set you cannot reach.

A live mirror of the help desk

Zendesk: Incremental Ticket Export returns every ticket created or changed since a point in time and hands back a cursor to resume from, which is exactly the shape a scheduled sync wants. With Zendesk: Incremental User Export and Zendesk: Incremental Organization Export alongside it, a Mattering database can hold a current mirror of tickets, requesters, and accounts for reporting and cross-app workflows.

People and organizations without the admin tab

Zendesk: Create User registers a requester (with an option to skip the verification email), Zendesk: Update User edits profiles, and Zendesk: Suspend or Unsuspend User is the fully reversible way to cut off a problem sender. Organizations get the same treatment: create, update, and exact-match search, with Zendesk: List Group Memberships showing which agents belong to which routing groups.

Macros you can preview before they touch a ticket

Zendesk: Preview Macro on Ticket uses Zendesk's own semantics: it returns the full ticket as it would look after the macro, without changing anything. An automation or an assistant can inspect the result, and only a follow-up Zendesk: Update Ticket with those fields commits it. Zendesk: List Macros and Zendesk: Get Macro cover discovery.

Example: Urgent tickets get an AI-drafted internal note and a review task

Built on the Automations canvas in a few minutes, or described in plain language to Matter and built for you.

1

Trigger: a schedule. There is no inbound Zendesk trigger in this connector, so polling on an interval is the honest pattern.

2

Zendesk: Search finds open urgent tickets with a query like type:ticket status:open priority:urgent, capped well below the search ceiling for a periodic sweep.

3

Zendesk: List Ticket Comments pulls each ticket's full conversation, oldest first, so the next step sees the whole thread.

4

An AI step drafts a suggested response from the conversation and the ticket's fields.

5

Zendesk: Update Ticket posts the draft as an internal note (a non-public comment), so nothing reaches the customer until a person decides it should.

6

A task opens in Mattering Productivity linking the ticket, and the agent reviews the draft there and sends the real reply from Zendesk.

Zendesk MCP questions, answered

What Zendesk operations are supported?

45 nodes on the Support API: tickets (list, get, create, update, soft-delete, restore, plus the deleted-tickets list and full comment history), unified search with count and an uncapped cursor export, users (list, get, create, update, search, suspend), organizations (list, get, create, update, exact-match search), groups and group memberships, macros with a preview-only apply, views (list, execute, full-ticket listing, count), read-only triggers and automations, tags (account list, append, remove), satisfaction ratings, ticket fields, attachment metadata, job status polling, and incremental exports for tickets, users, and organizations. Deliberately not shipped: bulk endpoints, ticket merge, permanent deletion, user deletion, attachment upload and redaction, writes to triggers, automations, macros, or views, and webhook management.

Does this work with Claude, ChatGPT, or other AI assistants?

Yes. Mattering exposes these nodes over MCP, an open protocol for connecting AI assistants to tools, so any MCP-capable assistant, including Claude and ChatGPT, can connect to your workspace and use them. Matter, the assistant built into the platform, uses the same nodes natively.

Can this connector permanently destroy data?

No node here erases anything outright. The one destructive node, Zendesk: Delete Ticket, is a soft delete: the ticket moves to Zendesk's deleted-tickets list, stays recoverable for up to 30 days, and only then is scrubbed. The connector ships the full recovery loop as Zendesk: List Deleted Tickets and Zendesk: Restore Deleted Ticket, and the permanent-delete endpoint is deliberately not built. User deletion is not offered at all, because Zendesk documents deleted users as not recoverable; Zendesk: Suspend or Unsuspend User is the reversible alternative, and lifting a suspension restores the user exactly.

Why does search stop at 1,000 results?

That is the Zendesk search API's own hard cap: at most 1,000 results per query, 100 per page, and page 11 returns an error, although the count field still reports the true total. Zendesk: Export Search Results exists for exactly this: it is cursor-paginated with no result cap, filtered to one record type per run, with cursors that expire after one hour and a recommended page size of 100. Use search for interactive-sized questions and the export for anything that might be large.

How do I add a comment or reply to a ticket?

Through ticket update, because the Zendesk API has no direct comment-create endpoint: comments are created by including a comment in a ticket update. Zendesk: Update Ticket has typed fields for the comment body and whether it is public. Public replies email the requester and fire business rules, so for automated or AI-drafted content the safe path is an internal note (public set to false), which only agents see. Zendesk: List Ticket Comments is read-only and returns the conversation.

Can a Zendesk event trigger a workflow?

Not as a trigger node in this connector: there is no inbound Zendesk trigger, and webhook management is deferred. The honest pattern today is polling. For keeping systems in sync, Zendesk: Incremental Ticket Export on a schedule is the right tool, since it returns everything created or changed since the last cursor. For sweep-style checks, a scheduled Zendesk: Search or view execution works; note that view results can be served from a cache up to 10 minutes old.

Why did my connection test fail saying Zendesk returned the anonymous user?

Because of a genuine quirk this connector refuses to paper over: Zendesk's show-self endpoint answers anonymous callers, so a request with a wrong credential can come back as a 200 with an anonymous user instead of a 401. A naive status-code test would call that success. Mattering's connection test therefore requires a real user identity in the response, and the anonymous case is reported as its own error. It means the subdomain, email, or token is wrong, or token access is disabled in Admin Center under Apps and integrations > APIs > Zendesk API > Settings.

What are the rate limits?

Plan-based, per Zendesk's published limits: 200 requests per minute on Team, 400 on Growth and Professional, 700 on Enterprise, and 2,500 on Enterprise Plus, with a High Volume add-on that raises qualifying plans to 2,500. Some endpoints carry their own caps, which the node descriptions state: 30 updates per 10 minutes per ticket, 10 incremental export calls per minute, 5 view executions per minute per view per agent, among others. Excess requests get a 429 with a Retry-After header; the connector does not retry silently, so put spacing or backoff in the automation where you control it.

45 nodes waiting on the canvas

Put Zendesk to work inside your workspace

Connect once, then use every node in visual automations and AI agent workflows across the Mattering platform.

All product names, logos, and brands are property of their respective owners and are used for identification purposes only. Use of these names, logos, and brands does not imply endorsement or affiliation. Zendesk is a product of its respective owner; Mattering is an independent platform.

Start Free Explore our Products