DHTMLX Kanban MCP server: cards, editor, and REST sync
A working DHTMLX Kanban board depends on several pieces fitting together correctly: card and editor fields need to share a matching key, the Toolbar runs as its own widget bound to the board, and server sync relies on a specific set of action handlers. Generated code holds up when it uses a matching key between card and editor fields, the Toolbar wired up on its own, and server sync built on today's action handlers, not a snapshot from training time.
The DHTMLX MCP server keeps the assistant working from current Kanban documentation instead of guesswork. Ask about binding editor fields to card fields, customizing the Toolbar, or connecting a RestDataProvider backend, and the assistant pulls the current reference before writing any code.
MCP endpoint
https://docs.dhtmlx.com/mcp
The DHTMLX MCP server covers all major DHTMLX products, not only DHTMLX Kanban. The same endpoint and configuration instructions work regardless of which DHTMLX component you are building with.
Kanban documentation via the MCP server
Under the hood, the server keeps a full index of the DHTMLX Kanban documentation, ready to serve requests like these:
- Looking up the current API for Kanban methods, events, and properties, including signatures and default values.
- Generating ready-to-run initialization code for both the Kanban board and the separate Toolbar constructor.
- Configuring card and editor fields (
cardShape,editorShape) and binding them by key, including combo, multiselect, color, date, and file field types. - Customizing card templates, context menus, and column headers with the documented template callbacks.
- Styling columns, rows, and cards with CSS variables and conditional CSS classes.
- Localizing the Kanban and Toolbar interfaces with built-in or custom locale objects.
- Loading, exporting, and modifying cards, columns, rows, links, and comments through the data API.
- Connecting a REST backend through
RestDataProvider, or setting up a multiuser board that syncs changes over WebSocket. - Integrating Kanban with other DHTMLX widgets such as Gantt and Scheduler, or with React, Vue, Angular, and Svelte.
What a Kanban prompt triggers in MCP
The DHTMLX MCP server answers Kanban questions through a Retrieval-Augmented Generation (RAG) pipeline layered on the Model Context Protocol (MCP), splitting each request between two workflows: Search, which pulls matching reference pages for the assistant to draw on, and Inference, which reads those pages itself and delivers a ready answer. Before any of that, the assistant first isolates which part of the request actually needs a documentation lookup and handles the rest from its own knowledge.
Take the prompt "How do I connect DHTMLX Kanban so that when a user adds a new card, it syncs automatically with my backend database?" as a walkthrough:
- The assistant identifies the part that needs documentation: how
RestDataProvidersyncs a new card to a server. - The server matches this to the server-integration documentation.
- Since answering means generating code, Search takes the request (a question with one correct answer would go to Inference instead).
- Search draws the matching pages from a vector index built on current Kanban documentation.
- The assistant receives those pages back as context.
- From that context, the assistant wires up
RestDataProviderfor theadd-cardoperation, then fills in the backend-specific request details from its own knowledge instead of guessing at the Kanban API.
Kanban code suggestions stay matched to the API as it works today.
Registering the MCP server per tool
Whether you prefer the terminal or a config file, either path points your tool at the URL below, and once it's registered, every Kanban board you build in that tool can reach it.
https://docs.dhtmlx.com/mcp
Setup steps for the most common tools follow below.
Claude Code
See Claude Code's own documentation for every supported way to register an MCP server.
To register the server from the command line, run:
claude mcp add --transport http dhtmlx-mcp https://docs.dhtmlx.com/mcp
Setting it up by hand? Add this entry to your .mcp.json:
{
"mcpServers": {
"dhtmlx-mcp": {
"type": "http",
"url": "https://docs.dhtmlx.com/mcp"
}
}
}
Cursor
For every MCP configuration path Cursor supports, see its official documentation.
Steps to add the server:
- Open Settings (
Cmd+Shift+Jon Mac,Ctrl+Shift+Jon Windows/Linux) - Go to Tools & MCP
- Click Add Custom MCP
- Paste the following config:
{
"mcpServers": {
"dhtmlx-mcp": {
"url": "https://docs.dhtmlx.com/mcp"
}
}
}
Google Antigravity
Antigravity 2.0
Full MCP integration details for Antigravity live in the official documentation.
These are the steps to complete for connecting DHTMLX MCP server with Google Antigravity:
- Open the command palette
- Type "mcp add"
- Select "HTTP"
- Provide the following values:
- Name:
dhtmlx-mcp
- URL:
https://docs.dhtmlx.com/mcp
Antigravity CLI
The related guide covers migrating from Gemini CLI to Antigravity CLI.
To connect the DHTMLX MCP server to Antigravity CLI, create mcp_config.json in one of these locations:
- Global:
~/.gemini/config/mcp_config.json - Workspace:
.agents/mcp_config.json
Add the following configuration:
{
"mcpServers": {
"dhtmlx-mcp": {
"serverUrl": "https://docs.dhtmlx.com/mcp"
}
}
}
Then run agy in the terminal.
ChatGPT
ChatGPT's official documentation walks through the full MCP connector setup.
Steps to configure the connector:
- Go to Settings → Apps & Connectors
- Click Advanced settings
- Enable Developer mode
- Return to Apps & Connectors and click "Create"
- Fill in the connector details:
- Name:
dhtmlx-mcp
- URL:
https://docs.dhtmlx.com/mcp
- Authentication:
No authentication
- Click Create
After you create the connector, ChatGPT pulls documentation from the MCP server during conversations.
For intensive coding workflows, other MCP-aware tools may be a better fit.
Other tools
Tools without a dedicated section here usually still support MCP, often labeled "Model Context Protocol" or "Context Sources" in settings. Add https://docs.dhtmlx.com/mcp there as a custom source.
What the MCP server does with your data
The MCP server is a hosted service: nothing runs on your machine, no files from your project are read, and no personal data is stored.
Debugging and service-improvement work may involve logging queries.
For stricter privacy requirements, organizations can request a commercial deployment with query logging turned off. Reach out to info@dhtmlx.com for details.
Kanban board prompts to borrow
The prompts below are grouped by the part of the API they target. Borrow one and adjust the specifics to your board.
Cards, columns, and swimlanes
How do I let users drag cards between columns and rows in DHTMLX Kanban? Use the docs.
How do I limit the number of cards allowed in a specific column or swimlane?
How do I duplicate a card and move the copy to a different row using the Kanban API?
Editor and field types
How do I bind a custom editor field to a card field so it appears on the card and in the editor?
How do I configure a multiselect field with user avatars in the DHTMLX Kanban editor?
How do I enable autosave with a debounce delay in the DHTMLX Kanban editor?
Toolbar and board controls
How do I add a custom search control to the Kanban Toolbar and bind it to the board?
How do I remove the undo and redo controls from the DHTMLX Kanban Toolbar?
REST and multiuser sync
How do I connect DHTMLX Kanban to a Go backend and load the initial data?
How do I customize how DHTMLX Kanban syncs a specific action to my backend?
How do I set up a multiuser DHTMLX Kanban board that syncs changes in real time over WebSocket?
Better results from Kanban prompts
- Name the target element. Distinguish between the board, a column, a row (swimlane), and a card (for example, "on the Kanban board" vs. "for a specific row" vs. "on a card in the Testing column"). A precise target helps the server retrieve the right reference pages.
- Specify the field type.
cardShapeandeditorShapesupport many field types (combo, multiselect, color, date, files). Name the type explicitly (for example, "a multiselect field with avatars") instead of "a field" so the assistant pulls the matching configuration reference. - Add "Use the docs" to your prompt. This phrase signals to the assistant that it should trigger an MCP lookup instead of answering from training data alone. It matters most for
RestDataProviderhandlers and field configuration, where the API is most likely to have changed. - State whether the operation is client-side or server-side. Kanban works with local data as well as a
RestDataProvider-backed REST API. Mention which one applies (for example, "using RestDataProvider" or "without a backend") so the assistant does not mix REST-specific calls into a client-only example.