Ask an AI app, and these tools do the work.
AI apps such as Claude and ChatGPT read, edit, and present the project you have open, using the tools SSHOW gives them. Here is what each tool does and what to say to use it, along with the tool names you see in the chat.
How the tools work
You don’t need to know the tool names. The AI app reads the conversation and calls the tools it needs on its own. For one request, it may use several tools in a row.
SSHOW tells the AI app to check which projects are open first, and to read the editing rules before its first edit. So when you say “Make the title on the first scene bigger,” it usually checks the project (list_projects), reads the rules (get_guide), reads the scene (get_scene), makes the edit (apply_actions), and takes a screenshot to see the result (get_screenshot).
The AI app’s edits and screenshots happen inside your editor, and its changes show up there right away. So keep the project open in the editor. If you connected with the cloud connector, four tools still answer from the last saved version when no editor is open: get_project_overview, get_scene, get_object, and find_objects. To connect, see Connect an AI app (MCP).
The tools differ slightly depending on how you connected. create_project exists only in the cloud connector, not in the desktop app’s MCP server. ChatGPT and Codex also get editing and presenting as separate tools, one per action. Those names are listed in the Kinds of edits and Presenting and testing interactions tables.
Find and read a project
| Tool | What it does | Ask like this |
|---|---|---|
| list_projects | Checks which projects are open in your editor right now. The AI app usually calls this tool first. If several are open, it works on the one you used most recently. To work on another one, name it. If none are open, it lists your 5 most recently edited projects with links to open them (cloud connector). | “Which SSHOW project is open right now?” |
| create_project | Creates an empty project in your account and gives you a link to open it. The AI app can edit it only after you open that link, and it counts toward your account’s project limit. Available only in the cloud connector. | “Create a new project called Product Launch.” |
| get_project_overview | Gets an overview of the project:the canvas size, the list of scenes with the number of objects in each, the current scene, the selected objects, the area visible on your screen, variables, fonts, the media in the project, the current edit mode, and the project notes. | “Summarize how this project is put together.” |
| get_scene | Reads one scene. It returns the scene’s name, presenter notes, background, transition, timeline, and interactions, plus the structure of the objects in it. If you don’t say which scene, it reads the current one. | “Tell me what’s on the third scene.” |
| get_object | Reads every setting of one object, plus its position and size on the canvas. | “Tell me the logo’s position and size.” |
| find_objects | Finds objects by name or text across all scenes. It looks inside groups and frames too, and returns up to 50 at a time. | “Find all the text that shows a price.” |
| get_fonts | Checks which fonts you can use for text:the list of fonts you can choose from, and the fonts already registered in this project. If you use a font that’s in neither, it shows as a system font, and the AI app is told so. | “Pick a font that suits the title.” |
| get_guide | Reads SSHOW’s guide to its editing rules. SSHOW tells the AI app to read it before its first edit, so the app usually calls it on its own. | No need to ask. |
Create and edit
| Tool | What it does | Ask like this |
|---|---|---|
| apply_actions | Creates, edits, and deletes objects, scenes, variables, and tables. Edits sent together in one go are undone with a single undo. When it makes a new scene, the editor switches to that scene. When a new object appears off screen, the view moves to it. In Animation mode with auto-keyframe turned on, edits that change values such as position, size, or opacity are recorded as keyframes at the playhead. Other edits, such as creating or deleting objects, apply as usual. For the edits it can make, see Kinds of edits. | “Add a title and a subtitle to the first scene.” · “Turn on transitions for every scene.” |
| import_asset | Imports an image, video, or audio file into the project from a link. The AI app places the imported file as an object in its next edit. The link must be an https address on a site that lets other sites load its files. | “Use the photo at this link as the background of the second scene.” |
| undo | Undoes the AI app’s last edit. If you’ve changed something since then, it doesn’t undo. To undo your own edits, press ⌘Z in the editor. | “Undo what you just changed.” |
| redo | Redoes an edit the AI app undid. If the next edit to redo is yours, it doesn’t redo. | “Redo what you undid earlier.” |
Kinds of edits
| Edit | What it does | Name in the chat |
|---|---|---|
| Create objects | Creates new rectangles, ellipses, paths, text, images, videos, audio, groups, frames, and tables. | create_object |
| Edit objects | Changes an object’s settings, such as its name, position, size, opacity, text, and whether it’s hidden or locked. | update_object |
| Change the style | Changes an object’s fills, strokes, and effects. On a scene, it changes the background. | update_object · update_scene (ChatGPT · Codex: set_style) |
| Add motion | Turns on transitions for scenes or objects, and adds keyframes and stops to the timeline. See Fine-tune transitions, Work with keyframes, and Stops and the work area. | update_object · update_scene (ChatGPT · Codex: set_motion) |
| Write interactions | Writes interaction sentences on scenes or objects. It doesn’t create or change requests. See How interactions work and Requests and variables. | update_object · update_scene (ChatGPT · Codex: set_rules) |
| Delete and duplicate | Deletes or duplicates objects. Deleting a group or frame also deletes what’s inside it. | delete_object · duplicate_object |
| Change the order | Brings objects to the front or back, or moves them one step forward or backward. It also moves objects into or out of groups and frames. Their position on the canvas stays the same. | move_object (ChatGPT · Codex: arrange_object) |
| Group and ungroup | Groups several objects together, or ungroups a group. | group_objects · ungroup |
| Convert to path | Turns a rectangle or ellipse into a path of the same shape. It can’t convert anything else, such as text, images, or an ellipse with an inner radius. If any of those are mixed in, none of the edits sent with it apply. For text, right-click it and choose Convert to Path. | convert_to_path |
| Table rows and columns | Adds or deletes rows or columns in a table. | insert_table_rows · insert_table_columns · delete_table_rows · delete_table_columns |
| Merge and split table cells | Merges several table cells into one, or splits merged cells apart again. | merge_table_cells · unmerge_table_cells |
| Manage scenes | Adds scenes, changes their name, presenter notes, and whether they’re hidden, and deletes or duplicates them. | create_scene · update_scene · delete_scene · duplicate_scene |
| Reorder scenes | Moves a scene to a different position. | move_scene (ChatGPT · Codex: reorder_scene) |
| Canvas size | Changes the canvas size that all scenes share. | set_scene_size (ChatGPT · Codex: set_canvas_size) |
| Manage variables | Adds variables, changes their name, type, and value, and deletes them. | create_variable · update_variable · delete_variable |
| Reorder variables | Moves a variable to a different position in the variable list. | move_variable (ChatGPT · Codex: reorder_variable) |
| Project info | Changes the project’s name, description, and notes. A new name also becomes the project’s name in Studio. | set_document (ChatGPT · Codex: update_document) |
Check and show on screen
| Tool | What it does | Ask like this |
|---|---|---|
| get_screenshot | Captures a scene or one object exactly as it looks now, for the AI app to see. The AI app uses it to check the result of an edit. It can capture any scene, but one object or one moment on the timeline only in the current scene. After capturing at a set moment, it puts things back as they were, and the captured image isn’t saved in the project. | “Check how the second scene looks.” · “Show me what it looks like at 1 second.” |
| set_active_scene | Moves your editor to another scene. The AI app uses it to show you where it made changes, or before it captures or selects objects in that scene. The document doesn’t change, so nothing is added to undo. | “Open the fourth scene.” |
| select_objects | Selects objects in the current scene to show you where they are. If it selects nothing, it clears the selection. | “Select the button you just changed so I can see it.” |
Presenting and testing interactions
The AI app starts and runs a presentation inside your editor window. During a presentation or an interaction preview, it sends input such as clicks to test interactions. For preview, see Edit and preview. For the triggers that respond to each input, see Triggers, actions, and conditions.
These tools don’t change the document, so there’s nothing to undo.
People with the play link open see the presentation too
Share live with viewers is on by default, so the scenes the AI app moves to also appear on the screen of anyone who has the play link open. To test on your own, turn this setting off in Play settings. Regardless of this setting, requests sent and URLs opened during a presentation or test really happen and can’t be undone.
| Tool | What it does | Ask like this |
|---|---|---|
| get_show_state | Reads how far the presentation has gotten. It sees the same things Presenter view shows: the current and next scenes, presenter notes, stops, what happens on the next advance, whether the screen is covered, and the elapsed time. | “Which scene are we on?” · “Read me this scene’s notes.” |
| control_show | Starts the presentation, advances it, jumps to a scene, pauses it, covers the screen, and ends it. The presentation opens inside the editor window. If you ask for full screen, it opens at the size of your screen and goes full screen the first time you click the presentation screen. The actions are in the table below. | “Start the presentation from this scene.” |
| simulate_event | Sends one input and sees which interactions ran, which were skipped and why, and what the screen looks like. The inputs it can send are click, double-click, long press, press down, release, pointer enter, pointer leave, swipe, key press, and message. Unlike a click in a real presentation, a click the AI app sends doesn’t advance the scene when no interaction receives it. To advance the scene, the AI app uses the Next action of control_show. | “Test whether clicking the Start button goes to the next scene.” |
| Action | What it does | control_show · ChatGPT · Codex |
|---|---|---|
| Start | Starts the presentation from the current scene or from the beginning. | start · show_start |
| Next | Advances like a clicker. If the scene has stops, it plays to each one in turn, then goes to the next scene. | next · show_next |
| Previous | Goes back one step, like a clicker’s back button. | previous · show_previous |
| Jump to a scene | Goes straight to the chosen scene. It doesn’t go to hidden scenes. | goTo · show_go_to |
| Pause and resume | Pauses the scene’s timeline where it is, or starts a paused timeline again. | togglePlayback · show_toggle_playback |
| Cover the screen | Covers the screen in black or white. Doing it again with the same color shows the scene again. | toggleCover · show_toggle_cover |
| End the presentation | Shows the “End of show” card to end the presentation. | endShow · show_end |
| Back to editing | Closes the presentation and returns to the editing screen. | exit · show_exit |
What AI apps don’t do
| Task | Details |
|---|---|
| Export to a file | No tool makes files such as videos or PDFs. Export them yourself as described in Export. |
| Sharing and invites | The AI app doesn’t set up play links or invite people to edit with you. See Share with a link and Edit together. |
| Delete a project | No tool deletes projects. The AI app can delete scenes and objects, and undo brings back anything it deletes. |
| Undo your edits | The AI app undoes only its own edits. To undo yours, press ⌘Z in the editor. |
| Create requests | The AI app doesn’t create interactions that send web requests, or change what they contain. But if the AI app rewrites the interactions and leaves out a sentence that has a request, that request is deleted too. If one disappears, bring it back with undo. To add a new request, write it yourself as shown in Requests and variables. |
| Read secret values in requests | When the AI app reads a request, header values and query values in the address are masked with asterisks. So an API key you put there isn’t visible to the AI app. Values in the request’s Body aren’t masked, so put secret values in Headers. |
| Projects in other accounts | The AI app sees only the projects of the account you allowed when you connected it. An editor signed in to another account isn’t visible to the AI app. |
What often comes up in the chat
-
Does it say no project is open?
The AI app’s edits and screenshots happen in an open editor. Open the project in an editor signed in to the account you allowed for the AI app, then ask again. -
Why does it ask for permission every time it uses a tool?
SSHOW marks each tool with whether it only reads, whether it does something hard to undo, and whether it reaches outside the project. AI apps read these marks and may ask for permission before tools that make changes. How you give permission differs from app to app, so follow that app’s instructions. -
It says it made the change, so why does the screen look the same?
The AI app may have edited a different scene from the one you’re looking at. Ask it to move to that scene, or choose that scene in the Scenes list. -
Did only some of what you asked for change?
apply_actions skips only the invalid edits and applies the rest. Ask the AI app why they were skipped and ask again, or fix those parts yourself. -
Why does it say it can’t undo when you ask?
If you’ve changed something since, the AI app doesn’t undo. Press ⌘Z in the editor to undo it yourself. -
Does it say it can’t import a photo?
This happens when the site blocks other sites from loading its files. Download the photo and add it to the editor yourself. -
Does the font look different from what you asked for?
A font that’s neither in the list of fonts you can choose from nor registered in this project shows as a system font. Ask the AI app to choose from the fonts get_fonts lists. -
Does it say it can’t create a new project?
The desktop app’s MCP server has no tool for creating projects, and the AI app can’t create one once your account’s project limit is reached. Create and open a new project yourself, or clear out projects you don’t use and ask again. -
Does it say it has no edit permission?
That AI app got read-only permission when it connected. Disconnect it in Settings, then connect again from the AI app.
Was this page helpful?
Updated 10/12/2026 · The videos on this page were recorded in the current editor.