Google Slides MCP Server

Generate and revise entire decks without opening the Slides editor.

The Slides service is built around the way presentations actually get made by agents: create_presentation starts the deck, batch_update_presentation applies a whole set of text, layout, and styling changes in one request, and get_presentation reads a single slide's element IDs and text, or its exact geometry and styles as compact JSON, for surgical edits.

render_page_to_image is the underrated tool here: it returns a slide as image content, so an assistant (or the human reviewing its work) can see what the deck looks like rather than reasoning blind about layout.

What you can ask

  • “Turn this outline into a ten-slide deck with a title slide and section headers”
  • “Update the footer and confidentiality notice across the entire presentation”
  • “Render the board deck slides so I can check the flow before the meeting”

All 6 Google Slides MCP tools

Expand any tool for its full parameter list. Tiers control how many tools the server registers: start lean with --tool-tier core and step up to extended or complete when a workflow needs more.

create_presentation Core Create presentations. 1
ParameterDescription
title The title for the new presentation. Defaults to "Untitled Presentation".
get_presentation Core Read a deck by action: "deck" lists its slides and layouts, "page" gives one slide's element IDs and all its text, "markup" returns one slide's elements as compact JSON with geometry and styles, and "layouts" lists every layout. 7
ParameterDescription
presentation_id requiredThe ID of the presentation.
action "deck" (default): title, URL, revision, and per slide its ID, title, layout, element counts and notes length, then the deck's layouts. "page": every element on one slide in z-order with its ID, text, table cells, image sources and speaker notes, no geometry or styling. "markup": one slide, or chosen elements, as JSON with boxes in points, text as paragraphs and runs, and only locally set styles. "layouts": every layout, grouped by master, with its placeholders and the slides using it.
page_id Slide (or any page) object ID, from "deck". Required for "page" and "markup".
targets For "markup": element IDs to return instead of the whole slide, plus "notes" for the speaker notes. An element inside a group comes back with page coordinates and its group ID.
raw For "markup": return the Slides API Page resource itself (EMU, transforms, text indices), limited to targets when given, for writing raw batch_update_presentation requests. Defaults to False.
offset For "deck": index of the first slide to list. "deck" lists 100 slides per call and a longer deck ends with the offset to call again with. Defaults to 0.
after For "layouts": a slide ID, or "start". Lists only the layouts of the master in use at that position, which are the only ones a new slide inserted there can use.
batch_update_presentation Extended Apply multiple slide updates in one request. 2
ParameterDescription
presentation_id requiredThe ID of the presentation to update.
requests requiredList of update requests to apply.
render_page_to_image Extended Render a slide or notes page to an image and return it as image content. 3
ParameterDescription
presentation_id requiredThe ID of the presentation.
page_object_id requiredThe object ID of the page/slide to render.
size Image width: "SMALL" (200px), "MEDIUM" (800px), "LARGE" (1600px), or "WIDTH2000_PX". Defaults to "MEDIUM", which reads titles and body text; use "LARGE" for tables, chart labels, and footnotes.
list_presentation_comments Complete List all presentation comments, including each comment's anchor so it can be matched to the slide element it refers to. —
manage_presentation_comment Complete Create, reply to, resolve, update, and delete presentation comments. Update and delete apply to top-level comments you wrote. —

Every service in one reference: full API documentation →

OAuth scopes

The server requests only the scopes for the services you enable. In read-only mode it requests just the readonly variants and skips registering write tools entirely.

  • presentations.readonly read decks and render slides to images
  • presentations create and edit presentations

Google Slides MCP questions

Can an AI actually verify what a slide looks like?
Yes. render_page_to_image returns any slide as image content. Multimodal assistants can inspect the rendered slide and fix overflowing text or broken layouts instead of guessing from the JSON structure.
How are bulk edits handled — one request per change?
No. batch_update_presentation accepts many operations in a single request, the same pattern the Slides API itself uses, so restyling a whole deck is one round trip rather than fifty.
Can it participate in deck review?
The Complete tier adds comment tools: list_presentation_comments to read feedback and manage_presentation_comment to reply, resolve, edit, or delete, so an agent can close the loop on review notes it has addressed.