MCP Tools

Growhouse provides MCP (Model Context Protocol) tools that allow AI assistants like Claude, Cursor, and others to interact with your canvases in real-time.

What is MCP?

Model Context Protocol is an open standard that enables AI applications to securely connect to external data sources and tools. With Growhouse MCP tools, AI assistants can:

Read canvas content - Access elements, text, and spatial structure
Create elements - Add sticky notes, text, images, videos, and social embeds
Update elements - Modify content, position, and styling
Understand context - AI-powered clustering detects moodboards, sequences, and relationships
Collaborate safely - Real-time awareness of other users editing

Available Tools

Core Tools

Tool	Description
`list_canvases`	List all canvases with MCP access enabled
`get_canvas_overview`	Get summary with element counts, bounds, and collaboration status
`find_elements`	Search by type, content, location, or collaboration status
`get_element_details`	Get full details including content, styles, and context

Content Creation

Tool	Description
`create_element`	Create sticky notes or rich text elements
`create_media_element`	Add images, videos, or social embeds (YouTube, TikTok, Instagram, etc.)
`update_element`	Modify content, position, or styles
`delete_element`	Remove an element from canvas

Intelligence & Collaboration

Tool	Description
`get_clusters`	AI spatial analysis - detect moodboards, comparisons, sequences
`get_collaboration_status`	Real-time info on online users, cursors, selections, locks
`resolve_natural_ref`	Convert human-readable refs like "sticky-top-left-1" to UUIDs

Element Types

Editable Elements

Type	Content Format	Best For
`sticky`	Plain text	Quick notes, brainstorming, ideas
`text-element`	Tiptap JSON	Formatted text, paragraphs, rich content

Media Elements (via `create_media_element`)

Type	Supported Sources
`media`	Direct image/video URLs (.jpg, .png, .gif, .mp4)
`embed`	YouTube, Instagram, TikTok, Twitter/X, LinkedIn, Facebook, Loom

Natural References

Instead of UUIDs, you can reference elements using human-readable IDs:

sticky-top-left-1      → First sticky in top-left quadrant
text-element-center-2  → Second text element in center
media-bottom-right-1   → First media in bottom-right

Use resolve_natural_ref to convert these to UUIDs when needed.

Smart Positioning

When creating or finding elements, use smart positioning:

Option	Example	Description
`quadrant`	`"top-left"`	Position in canvas region
`near_element`	`"sticky-top-left-1"`	Place 50px right of another element
`x, y`	`{x: 100, y: 200}`	Exact coordinates

Real-Time Collaboration

All read tools support collaboration_context to see:

Who is online viewing the canvas
Which elements are locked for editing
Which elements are selected by others
Cursor positions of collaborators

This prevents conflicts when AI and humans work together.

Use Cases

Content Generation

"Add a sticky note with 'Q1 Goals' in the top-left,
then add three more stickies below it with our priorities"

Media Curation

"Add this YouTube video to the moodboard section"
"Embed these Instagram posts near the inspiration cluster"

Canvas Analysis

"Analyze this canvas and tell me what themes you see"
"What are the main clusters and how do they relate?"

Safe Collaboration

"Show me which elements are available to edit right now"
"Update the sticky that no one else is working on"

Getting Started

Setup Guide - Connect to your AI assistant
Canvas MCP Reference - Complete API reference

MCP Tools Overview