The Google Slides MCP server that builds the client deck

Google Slides MCP connects Google Slides to AI assistants through the Model Context Protocol. You describe the deck, and Claude, ChatGPT, or another MCP client creates presentations, fills templates, inserts charts and tables, and updates existing decks, with every change held for your approval.

No. Google publishes official MCP servers for some products, but not for Slides. The options are community open-source servers, connector platforms like Zapier or Composio, and managed layers like Markifact.

Two things. First, the data: marketing decks present numbers, and with Markifact the numbers come from Google Ads, GA4, Meta, and TikTok in the same conversation that builds the deck. Second, linked Sheets charts: Markifact inserts and refreshes them, which most alternatives skip, and it is the difference between rebuilding a deck monthly and refreshing it.

Yes. The slides_create_presentation operation starts the deck, and slides_add_slide, slides_insert_text, slides_add_image, and slides_create_table build it out. The draft always waits for your approval in Markifact before anything is created.

Yes, and it is the recommended workflow. The AI duplicates slides from your template with slides_duplicate_slide and swaps the placeholders with slides_replace_text, so your fonts, colors, and layouts survive intact while the content updates.

The slides_add_sheets_chart operation inserts a chart connected to a live Google Sheet, and slides_refresh_sheets_chart updates it when the data changes. Build the monthly report deck once, and from then on the charts refresh instead of being recreated.

From the same MCP connection. Pull Google Ads spend, GA4 traffic, or Meta performance in one message, and put it into the deck in the next, no exports and no copy-paste between tools.

Yes. The slides_get_slide and slides_list_elements operations read slide content, which lets the AI audit a deck for leftover placeholders, summarize a long presentation, or check what it is about to change.

Yes. The slides_download_presentation operation exports the deck so it can be attached to the client email or archived, all from the same session.

Markifact separates reads from writes. Reading a deck runs instantly; anything that creates or changes slides stops and waits for your explicit approval in Markifact, so a template never gets overwritten without you seeing the change first.

No. Running a Slides MCP yourself means creating a Google Cloud project and managing OAuth credentials. With Markifact you sign in with Google once and the hosted MCP server handles the rest.

Yes, and it is the workflow this page exists for: duplicate the template, pull the month's numbers from your ad platforms, replace the placeholders, refresh the linked charts, and approve. What used to be an afternoon becomes a review.

Markifact uses a secure OAuth-based connection scoped to your Google account. Your tokens are never exposed to the AI client, and deck content is processed in real time without retention.