Notiondesk MCP Search makes your published help center available to compatible AI assistants through the Model Context Protocol (MCP).
Once connected, AI assistants such as ChatGPT, Claude, Cursor, and Codex can search your documentation and read the relevant published content when answering questions.
MCP Search is public and read-only. It does not give AI assistants access to your Notiondesk workspace or allow them to modify your help center.
What is MCP Search?
The Model Context Protocol, or MCP, is an open standard that lets AI assistants connect to external tools and sources of information.
Notiondesk MCP Search provides an MCP connection specifically for your published help center.
Instead of relying only on training data, web search, or content manually pasted into a conversation, a connected AI assistant can search your current documentation when it needs an answer.
For example, a reader could ask:
How do I configure automatic translations?The AI assistant can search your Notiondesk help center, find the relevant articles, and use that documentation to answer.
What can AI assistants access?
MCP Search can give connected AI assistants read access to:
- Published help center articles
- Collections
- Changelog entries
Only content that is public and published through your help center is available through MCP Search.
MCP Search is read-only
MCP Search cannot modify your help center.
A connected AI assistant cannot use MCP Search to:
- Create articles
- Edit content
- Publish or unpublish articles
- Trigger syncs
- Start translations
- Change settings
- Delete data
- Access private workspace information
MCP Search is designed for knowledge retrieval, not workspace management.
If you want authorized teammates to manage Notiondesk from an AI assistant, use Workspace MCP instead.
Who should use MCP Search?
MCP Search is useful when you want people or AI systems to access your published documentation outside the Notiondesk help center itself.
Typical use cases include:
- Customers asking product questions from ChatGPT or Claude
- Developers searching your documentation from Cursor, Claude Code, or Codex
- Internal teams accessing published documentation from their AI assistant
- Making product knowledge available to AI agents
- Giving readers another way to search your help center
Readers do not need access to your Notiondesk workspace to use MCP Search.
No API key or Notiondesk login is required
MCP Search for a public help center does not require an API key or Notiondesk account.
Each enabled help center has its own MCP connector that can be added to compatible AI assistants.
This makes it possible to share access to your published documentation without exposing private workspace credentials.
Multilingual help centers
MCP Search works with published translated content.
When your help center is available in multiple languages, an AI assistant can search and use the appropriate translated documentation when answering readers.
This lets the same MCP connection serve users asking questions in different supported languages.
Supported AI assistants
Notiondesk provides connection instructions for:
- ChatGPT
- Claude
- Claude Code
- Cursor
- VS Code
- Codex
Other AI clients can also connect when they support compatible remote MCP servers.
See Add your help center as a connector in Claude, ChatGPT, or Cursor for installation instructions.
MCP Search vs Workspace MCP
Notiondesk provides two different MCP connections.
MCP Search is for accessing published knowledge.
Workspace MCP is for authorized teammates who want an AI assistant to work with internal Notiondesk information and supported workspace actions.
A simple way to distinguish them is:
MCP Search helps AI assistants consume your knowledge. Workspace MCP helps your team operate Notiondesk.
See Workspace MCP vs MCP Search for a detailed comparison.
Availability
MCP Search is available on the Pro plan and above.
It must be enabled for the help center before readers or teammates can connect it to an AI assistant.
See Enable MCP Search for your help center for setup instructions.