Connect your AI
Modula exposes a Model Context Protocol (MCP) server that any compatible AI assistant can connect to. Once connected, the AI gains access to the full Modula toolset — it can look up the block types, build pages block by block, and deliver production-ready interactive lessons — all from a natural-language prompt.
The MCP server
The Modula MCP server is available at:
https://modula.dev/mcpIt uses the Streamable HTTP transport — a single HTTPS endpoint that accepts POST requests. This is the modern MCP transport and is supported by all major AI clients.
How authentication works
The server uses OAuth 2.1 with PKCE. The first time your AI calls a Modula tool in a session, it opens a browser window to complete the Modula login flow. Once you sign in and grant permission, the connection is established and persists for that session.
You only authenticate once per client — subsequent sessions use a refresh token automatically.
What the AI can do once connected
| Tool | What it does |
|---|---|
get_schema_instructions | Returns the instructions for building a kind of page (a lesson, a presentation, …) from blocks |
list_types | Returns the block types available to you, with their variants |
design_blocks | Describes the blocks the AI is about to write (a text, a gallery, panels holding other blocks, …) and returns the exact form of their content |
edit_blocks | Creates or edits those blocks: returns updated data and stable ids and opens them in an editor |
export_blocks | Exports the updated blocks as one HTML fragment per block, in order, plus the shared script; the AI composes the page. For an LMS it also returns the SCORM manifest and tracking script |
open_document | Reads a Modula HTML file’s contents and returns its ordered blocks, ids and metadata so the AI can resume editing |
save_document | Returns a complete HTML file with editable source and the script included, ready to save and reopen in the Modula editor |
get_theming_instructions | Returns the colour tokens of a page, which the AI sets as CSS custom properties |
ai_respond | Answers a request for AI help made inside a block editor, using your assistant’s model |
submit_support_ticket | Sends a support ticket on your behalf |
To continue working on a Modula document, provide its HTML file contents to your assistant. After editing, the assistant can return a complete file for you to save or download. These document tools do not read local file paths, fetch URLs or store documents on the server.
Supported AI clients
| AI client | Connection method |
|---|---|
| Claude.ai | Settings → Customize → Connectors |
| Claude Design | Inherits connectors from Claude.ai |
| Claude Code | CLI command or ~/.claude.json config |
| ChatGPT | Settings → Apps & Connectors (plans and workspaces that allow custom apps in Developer Mode) |
| Codex | codex.json config file |
| Cursor | Settings → Features → MCP |
| GitHub Copilot | VS Code settings or mcp.json config |
Select your AI client from the list above or from the sidebar for step-by-step setup instructions.