Connecting AI coding agents to the TinyMCE documentation

An AI coding agent works from its training data unless it is given something better. This page describes the ways to put the current TinyMCE documentation in front of an agent: a documentation Model Context Protocol (MCP) server, markdown versions of every page, and the llms.txt summaries. Each method applies to any TinyMCE installation, whether it is served from the Tiny Cloud or self-hosted.

To add TinyMCE to a project with an AI coding agent, see the Tiny Cloud AI quick start guide.

This page covers giving external coding agents access to this documentation. It does not describe the in-editor AI features. For content generation inside the editor, see TinyMCE AI.

Connect the documentation MCP server

The hosted TinyMCE documentation MCP server exposes semantic search across the published documentation, returning relevant extracts with links to their source pages. Any client that supports MCP over streamable HTTP can connect to it. This server supplies documentation to an external coding agent, and is separate from the MCP tool calling available to the TinyMCE AI service, described in MCP and web integrations.

Endpoint: https://tinymcedocs.mcp.kapa.ai

Authentication: The server is protected by OAuth and requires a one-time sign-in. The first time the agent calls the server, a browser window opens to complete sign-in, after which the agent stores the resulting token. In Claude Code, if no sign-in prompt appears, run /mcp, select the server, and then select Authenticate.

Each agent reads its own configuration file, and the files are not interchangeable. Claude Code reads .mcp.json and expects an mcpServers object, while Visual Studio Code reads .vscode/mcp.json and expects a servers object. An agent given another agent’s file or key name reports no error and exposes no tools from the server. Expand the entry below that matches the agent in use.

Install from the documentation site

Every page of this documentation includes an Ask AI widget. Open the widget and select Use MCP in the modal header to open the Connect to AI Tools menu, which provides copy-to-clipboard install commands, one-click installs for supported editors, and the raw server URL.

The entries below describe the equivalent manual configuration.

Manual configuration by agent

Each agent below is configured independently. Expand the entry for the agent in use.

Claude Code 
claude mcp add --transport http --scope project tinymce-docs https://tinymcedocs.mcp.kapa.ai

To configure the server manually, add it to .mcp.json:

{
  "mcpServers": {
    "tinymce-docs": {
      "type": "http",
      "url": "https://tinymcedocs.mcp.kapa.ai"
    }
  }
}
Cursor

Add the server to .cursor/mcp.json:

{
  "mcpServers": {
    "tinymce-docs": {
      "url": "https://tinymcedocs.mcp.kapa.ai"
    }
  }
}
Codex 
codex mcp add tinymce-docs --url https://tinymcedocs.mcp.kapa.ai

To configure the server manually, add it to ~/.codex/config.toml:

[mcp_servers.tinymce-docs]
url = "https://tinymcedocs.mcp.kapa.ai"
startup_timeout_sec = 20
tool_timeout_sec = 60
enabled = true
Visual Studio Code

Add the server to .vscode/mcp.json:

{
  "servers": {
    "tinymce-docs": {
      "type": "http",
      "url": "https://tinymcedocs.mcp.kapa.ai"
    }
  }
}
OpenCode

Add the server to opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "tinymce-docs": {
      "type": "remote",
      "url": "https://tinymcedocs.mcp.kapa.ai",
      "enabled": true
    }
  }
}
Other agents

For any other agent, point its MCP client at the streamable HTTP endpoint and complete the sign-in described above. Agents that accept a remote MCP server without a dedicated configuration file, such as Claude and ChatGPT, can connect using the endpoint directly:

https://tinymcedocs.mcp.kapa.ai

Connect Context7

Context7 is a general-purpose documentation MCP server that also indexes the TinyMCE documentation. Where the TinyMCE documentation MCP server returns extracts from the published site, Context7 serves version-specific library documentation across many projects from a single server, which suits an agent already using it for other libraries.

Library: tinymce/docs at context7.com/tinymce/tinymce-docs

Configure Context7 following its own installation instructions, then reference the tinymce/docs library in prompts so the agent resolves TinyMCE questions against it.

Read the documentation as markdown

Agents can read the documentation efficiently without the MCP server. Every documentation page is also published as markdown, which is more token-efficient than the rendered HTML page. Append index.md to any documentation page URL to retrieve it:

https://www.tiny.cloud/docs/tinymce/latest/basic-setup/index.md

Markdown is available for every documented version of TinyMCE, not only the latest release. Replace latest with a version segment to retrieve the markdown for an earlier version:

https://www.tiny.cloud/docs/tinymce/7/basic-setup/index.md

Each page also includes a Copy Markdown button next to the page title, which copies the markdown URL of the current page to the clipboard for pasting into an agent prompt.

llms.txt files

Two plain text files summarize the documentation set for agents that accept a single bulk reference. Both files index the latest release:

File Contents

llms.txt

A concise overview of TinyMCE, with key links and code examples.

llms-full.txt

A full index of the documentation pages and their URLs.

Feedback

Report inaccurate or outdated documentation by opening an issue in the tinymce-docs repository. Agents are encouraged to file these when the documentation contradicts what actually worked.