> ## Content Index > Fetch the complete content index at: https://internalnote.com/llms.txt > Use this file to discover other available public pages before exploring further. # Universal Connector for Zendesk Knowledge Graph sources - URL: https://internalnote.com/universal-connector-for-zendesk-knowledge-graph-sources/ - Published: 2026-09-10T08:10:54.000Z - Updated: 2026-09-10T08:10:54.000Z - Description: Knowledge Connectors bring external content into the Zendesk Knowledge Graph, so the same information can support Help Center search, AI Agents and Agent Copilot. Dedicated connectors make common sources straightforward to connect. The Universal Connector extends that to any API. - Author: Thomas Verschoren - Tags: Knowledge Most organisations already have more knowledge than their support operation can use. Product troubleshooting lives on a Help Center. The documentation itself lives in a Notion doc. Engineering guidelines live in Confluence. Policies sit in SharePoint. Product or marketing teams manage content in Contentful, while an internal application exposes API documentation directly via a JSON payload. The problem is not that the information doesn't exist. But the issue is that these sources are often separated at exactly the moment they need to work together. A customer asks a question but the most accurate information is somewhere the AI Agent can't access. The usual results are either a non-answer or an human agent answering the question after an escalation. A Patchwork solution might be copying those answers into the Help Center. Which leads to fragmented data, copies of the same information living across tools, and worse content that grows stale or out of sync with reality. Zendesk Knowledge Connectors address that gap by making external knowledge available inside Zendesk's own knowledge layer. They do not replace the systems that own the content. They make the content discoverable where service happens, while leaving its source and ownership in place. This is the next step in the shift from a Help Center as a destination to knowledge as infrastructure. The Help Center remains an important source, but it is no longer the boundary of what Zendesk can know. [Enriching Zendesk’s Knowledge Graph With External SourcesZendesk’s Knowledge Graph turns knowledge into a connected platform layer. This article shows how Federated Search and Knowledge Connectors bring content from Help Centre, websites and internal tools into one graph, then surface, measure and improve it across channels.![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/icon/internalnote_color@2x-0f7f3145-1608-48c4-9f42-fa069be1a8fc.png)Internal NoteThomas Verschoren![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/thumbnail/Banner---2026-01-Agentic-1-110364c0-f83c-4459-a979-f8854cb6e154.png)](https://internalnote.com/knowledge-connectors/) # Knowledge Graph Zendesk now has three practical routes for getting content into the Knowledge Graph. Your **Help Center** is the native source. This is where support teams create and govern customer-facing and internal articles directly in Zendesk and is often the starting point for many a Zendesk user. Federated Search, or **web crawlers**, is the route for public content that lives on a website, blog, shop or documentation site. Zendesk can crawl those destinations, index their content, and make them available without the need to copy them over into Help Center articles. **Knowledge Connectors** are the route for content held in another knowledge system or business application. These often have specific data structures or permission models that are not ideal for web crawlers. A prebuilt connector syncs these sources into Zendesk as external content, where it can be used alongside Help Center content. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Knowledge-Graph-2.png) Once external content is connected, it can be used for both classic Help Center search as well as Generative Answers. It's available in the Knowledge panel in Agent Workspace, in AI Agent responses, and as a source for Agent Copilot auto-assist. The content enters through one connection, but it can support several resolution experiences. This shared Knowledgh Graph is key. A team does not need one copy of a policy for the Help Center, another for the AI Agent and a third for agents. It gets a well-governed source that each experience can use, with centralised analytics as the result. 💡 It is worth separating Knowledge Connectors from the connectors used in Action Flows. They solve different problems. An Action Flow connector retrieves or changes `data` in an external system as part of an execution. A Knowledge Connector ingests `content` so that it can be retrieved by search and used to ground an answer. One performs work; the other makes knowledge available. # Knowledge connectors The prebuilt Knowledge Connectors focus on the systems most organisations use as internal knowledge repositories: Confluence, Notion, SharePoint, Document360 and Box. The current Zendesk reference lists dedicated connectors for over a dozen popular tools, ranging from Confluence and Google Driveup to Salesforce, SharePoint or even Zendesk content itself. The connectors are source-aware, so the setup can expose the relevant account, site, workspace, collection, folder, space, category or locale immediately, without asking an administrator to describe the source's data structure. [Connecting external knowledge sources to your Zendesk accountYou can connect an external knowledge source to your Zendesk account to make the content availablewherever external content is used. Each knowledge connector has a connector-specific authentication…![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/icon/favicon-aa0fc79c-ad90-444c-80df-380910a8a9f9.ico)ZendeskZendesk![](https://support.zendesk.com/static.zdassets.com/logo.png)](https://support.zendesk.com/hc/en-us/articles/9796599390874-Connecting-external-knowledge-sources-to-your-Zendesk-account?ref=internalnote.com#topic%5Fa4w%5Fks2%5Fk3c) Knowledge Connectors also reaches beyond pages. Depending on the connector, Zendesk can ingest structured files such as Markdown or PDF (currently in EAP). Structured, editable content is still preferable where it exists, because it is easier to review and update, but the connector model is no longer limited to conventional web pages. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Sources-2.png) An overview of the current set of supported connectors. # The universal connector The practical rule is simple: use a dedicated connector when one exists. It understands the source and usually gives an admin the shortest path from authentication to a scoped sync. Now, let's imagine a company with an internal policy portal built for its own business. It contains useful, article-like content, but the tool is built internal, so there's no prebuilt connector. The portal does have a REST API. Each record contains an identifier, a heading, the body of the policy, and a path to the source page. You used to have three options to get this content into Zendesk: 1. You can copy the policies into the Help Center and accept the maintenance burden of keeping the data up to date 2. You can ask developers to build a synchronisation service, leveraging the [Federated Search API](https://internalnote.com/federated-search/). And then find someone to monitor credentials, schema changes, pagination and failures. 3. Or it can leave the portal disconnected and accept that AI Agents and agents will answer without a relevant source. Now, there's a fourth solution: the **Universal Connector**. The Universal Connector allows you to sync *any* data sources that has an API endpoint. An administrator supplies the API endpoint and API key. Zendesk connects to the API and shows a preview of the response data. You can map the returned fields to Zendesk elements, and tell Zendesk how to retrieve subsequent pages. After which Zendesk will use that API to sync its contents on a regular basis adding, removing or updating the synced data. The source remains the system of record, but its content becomes part of the same searchable knowledge layer. The Universal Connector is not a universal integration builder however.. It is very specific tool that focused on getting structured content data into the Knowledge Graph. If you want more generic data like retrieving a specific order status, you need to use [Custom Actions](https://internalnote.com/more-powerful-action-flows/) and Action Flows. The source must expose a REST API that accepts `GET` requests, return `JSON`, authenticate with an API key, include the content and required metadata in each record, and expose a pagination pattern Zendesk can configure. Synchronisation is a full refresh every 24 hours, with the option to trigger a manual resync. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Generic-1-1.png) ## Adding a new universal source The Universal Connector is easier to understand with a quick demo. For a demonstration, I use [The Movie Database (TMDB) API](https://www.themoviedb.org/settings/api?ref=internalnote.com) and import its list of popular movies as external content. In Knowledge admin, go to **Manage articles > External content > Connections**. Under **Available sources**, select **Universal Connector**, review the connection message and continue. Give the connection a name such as `Popular movies`. Enter the API URL and the API key. If required by the endpoint, add query parameters. The Universal Connector supports up to five query parameters, so this is where you can set values such as `language=en-US` or an initial page value when the API expects one. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Generic-2-1.png) Zendesk uses the endpoint and credentials to validate the connection and retrieve a sample response. The fields returned by that response become available for mapping. ## Mapping the response When we retrieve movie popular movies from TMDB , we get those in the following format: ```json { "page": 1, "results": [ { "backdrop_path": "/qeQJx07rK2xm8SD2sJxFKhE7gs0.jpg", "id": 969681, "title": "Spider-Man: Brand New Day", "original_language": "en", "original_title": "Spider-Man: Brand New Day", "overview": "Fighting crime full-time as Spider-Man in a world that doesn't remember him—and the pressure of seeing his old friends move on without him—sparks a change in Peter Parker he may not have the power to control. But that transformation might also be the only thing that can stop a shocking new threat to the city and those he loves - a powerful villain no one can even see.", "release_date": "2026-07-29", }, /*...*/ ], "total_pages": 58800, "total_results": 1175995 } ``` We can then map those fields to Zendesk's required fields: | Zendesk field | TMDB field | Purpose | | ----------------- | -------------- | ----------------------------------------------------------------------------------------------------------- | | Unique identifier | id | Keeps each movie record distinct across syncs | | Title | title | Becomes the external article heading | | Body | overview | Supplies the content used by search and AI features | | URL | backdrop\_path | My API did not have a relevant URL field. But its useful if you want to link to the original source article | ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Generic-3.png) The Universal Connector also exposes mappings for **Last updated** and **Locale**. Map those only when the source returns appropriate values. For this demo, `language=en-US` is a request parameter, not a per-record locale field, and `release_date` is not the same thing as the date on which the record was updated. So I opted to leave those blank. The URL field deserves attention. A source may return a relative path such as `/movie/12345` or an identifier such as `12345`, without returning the host name. These currently can be added as a `URL` value, but won't link to a valid webpage. Currently, you can't concat together a full URL, `https://domain,com` \+ `{{movie_url}}`, but that's planned to be supported in the future. Your API might require pagination so that the import tool can loop through all available content. You can use paginated pagination, or cursor-based pagination. For each, you can select the right parameter in the returned payload. # Activating sources Connecting a source and activating it are separate steps. The first puts content into Zendesk. The second decides which features may use it. ## Help Center and Generative Search You can add the connection as an external search source for the Help Center. With Generative Search enabled, Zendesk can use the connected records to produce a direct answer above the usual search results, with the external source available for context. Otherwise, you'll get the classic list of blue links the user can click on. You can do this by first making the source visible to everyone (or a segment) in the Connections page of Knowledge Admin. Then you need to go to Search Sources under Settings, and add our newly imported source as a Knowledge Source. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-1.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-2.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-3.png) Once enabled, the Help Center will use these newly imported items to start answering questions. In the example below I use the new [Lisbon](https://internalnote.com/introducing-lisbon/) theme and its conversational experience, as well as the search results in the classic Copenhagen theme, combining Generative Search and search lists. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Result-1.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Result-2.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Result-3.png) ## AI Agents External content can be used by Zendesk AI Agents. Add the connected source in the AI Agent's knowledge or content settings, then test questions that should be answered from the imported records. [Connecting knowledge sources to power generative replies in AI agentsKnowledge sources are the information your AI agent uses to create AI-generated answers to questions from your customers. Adding knowledge sources to your AI agent lets it generate answers to help…![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/icon/favicon-39865211-76a2-4a5e-8321-e02a6c88338b.ico)ZendeskZendesk![](https://support.zendesk.com/static.zdassets.com/logo.png)](https://support.zendesk.com/hc/en-us/articles/8357749301658-Connecting-knowledge-sources-to-power-generative-replies-in-AI-agents?ref=internalnote.com) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-AI-2.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-AI-1-1.png) Once the AI Agent finishes indexing the external source, it'll start replying leveraging this new content: ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Widget-2.png) ## Agent Copilot For Agent Copilot, add the connection as a source for auto-assist. With the new [Guided Onboarding](https://support.zendesk.com/hc/en-us/articles/11145312935066-Announcing-new-guided-onboarding-for-Copilot-auto-assist?ref=internalnote.com) this step becomes an easy toggle during the setup process. Auto assist can then use the connected content to propose a reply, while the human agent can open the source and verify the suggestion before sending it. ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-Copilot-1.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-Copilot-2.png) ![](https://storage.ghost.io/c/ca/c0/cac0c82b-bc4d-4404-9eb4-9cc75e8d045e/content/images/2026/09/Activate-Copilot-3.png) The same source can therefore appear in three different forms: as a Generative Search answer for a customer, as grounded context for an AI Agent, and as a suggested response for a human agent. The content is shared, but the resolution experience is different. # Knowledge governance The connector makes ingestion easier, but it does not remove the need for knowledge governance. First, scope the credential carefully. The Universal Connector, or any connector for that matter, can retrieve anything that the configured API key or authentication method can retrieve, and Zendesk will not apply the source's individual record permissions afterwards. Use a credential designed for the intended knowledge set and set the connection's viewer permissions deliberately. 💡 This is what makes Zendesk's Employee Service experience unique. The new Employee Service Agents, based on their acquisition of Unleash, does support content filtering and authentication based on the **end-user* asking the question. Second, especially if this is a in-house built tool, treat the API and its source response as a contract. Each record should have a *stable* *unique* *identifier*, a useful title and readable body content. If the API sometimes omits required fields, changes its response shape or returns an error instead of an empty page, the sync may not produce the result you expect. Third, plan for freshness. A full refresh runs every 24 hours, although an administrator can manually re-sync the connection. This is suitable for policy and documentation sources that change periodically. It is not a real-time feed for operational data. Finally, connect only content that should be used. A larger index is not automatically a better one. Duplicates, outdated pages and loosely related records make retrieval less reliable and make it harder to understand which source should be trusted. The goal is not to import every available response. It is to expose the smallest useful set of governed knowledge. # With great power... Dedicated connectors make the popular platforms easier to connect. The Universal Connector covers the long tail: internal systems, niche knowledge bases and custom applications that have a usable REST API but no dedicated integration. Its value is not that every API suddenly becomes a Zendesk data source. Its value is that article-like content can cross the boundary into the Knowledge Graph without being copied, rebuilt or hidden from the experiences that need it. Once the source is connected and governed, Help Center search, Generative Search, AI Agents and Agent Copilot can all draw from the same external knowledge. That is the role of Knowledge Connectors in the Zendesk ecosystem: not another place to write content, and not another way to execute actions, but a controlled path for the knowledge an organisation already owns to become part of every resolution. So, what are you connecting tomorrow? ## Sign up for Internal Note Turning Zendesk into practice. – A weekly newsletter written by Thomas Verschoren. Subscribe Email sent! Check your inbox to complete your signup. No spam. Unsubscribe anytime.