Developers & AI

Connect Claude or Cursor with MCP

Point your AI client at the Ettajer MCP endpoint and authorize with OAuth.

  1. Model Context Protocol (MCP) lets Claude and Cursor call Ettajer tools over HTTP. Production endpoint: https://www.ettajer.com/api/v1/mcp — method POST, JSON-RPC 2.0, header Authorization: Bearer <access_token or etsk_live_…>.

  2. Prerequisites: an OAuth app in the developer console with the correct redirect URIs, and either a completed OAuth authorize flow or an API key on that app.

  3. In Claude: add a custom MCP connector / server pointed at the Ettajer endpoint, then complete OAuth when prompted. In Cursor: add an MCP server with the same URL and OAuth (or paste a Bearer API key if your setup supports it).

  4. Smoke-test after connect: call initialize, then tools/list and resources/list. You should see theme and catalog tools and resources such as ettajer://store , ettajer://theme-schema , ettajer://products , ettajer://collections , ettajer://navigation .

  5. Recommended agent workflow: get_context (follow workflow.next) → get_theme_schema → create_theme only if no draft exists → apply_theme_batch or update_section → preview_theme → ask the merchant to publish (or publish_theme only with themes:publish).

  6. Useful tools include get_store, get_context, get_products, get_product, get_collections, get_themes, get_theme, get_theme_schema, create_theme, update_theme, create_page, update_page, delete_page, create_section, update_section, delete_section, get_media, upload_media, get_navigation, update_navigation, preview_theme, publish_theme.

  7. preview_theme returns a short-lived signed previewUrl so you can open the draft without a merchant browser cookie. publish_theme validates then applies live — keep it off for AI clients unless you explicitly granted themes:publish.

  8. Principle: AI designs presentation (theme, layout, copy structure, media). Ettajer keeps cart, checkout, orders, and payments. Do not invent commerce endpoints for agents.

  9. Never put client secrets, API keys, or access tokens in documentation, screenshots, or chat transcripts. Rotate compromised credentials in the console immediately.

  10. Docs: /developers/mcp , /developers/quickstart , /developers/ai-system-prompt for the agent system prompt.

Was this article helpful?

Contact support