loading…
Search for a command to run...
loading…
Enables AI agents to read, write, and search local tldraw (.tldr) files, providing a persistent visual scratchpad for diagramming and note organization. It supp
Enables AI agents to read, write, and search local tldraw (.tldr) files, providing a persistent visual scratchpad for diagramming and note organization. It supports full CRUD operations on canvas shapes and metadata management for local canvas files.
MCP (Model Context Protocol) server for managing local tldraw canvas files (.tldr).
Existing tldraw MCP servers let AI draw on an in-memory canvas. This project is different — it reads, writes, and searches local .tldr files on disk, making tldraw a persistent visual scratchpad that AI agents can programmatically update.
.tldr files.tldr files with metadatanpx @talhaorak/tldraw-mcp
git clone https://github.com/talhaorak/tldraw-mcp.git
cd tldraw-mcp
bun install
bun run start
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"tldraw": {
"command": "npx",
"args": ["@talhaorak/tldraw-mcp"],
"env": {
"TLDRAW_DIR": "/path/to/your/tldraw/files"
}
}
}
}
Add to your OpenClaw config:
mcp:
servers:
tldraw:
command: npx
args: ["@talhaorak/tldraw-mcp"]
env:
TLDRAW_DIR: /path/to/your/tldraw/files
Or install as a skill:
curl -fsS https://raw.githubusercontent.com/talhaorak/tldraw-mcp/main/SKILLS.md | bash
| Variable | Default | Description |
|---|---|---|
TLDRAW_DIR |
~/.tldraw |
Base directory for .tldr files |
tldraw_readRead a .tldr file and return its parsed content.
tldraw_read({ path: "notes.tldr" })
tldraw_writeWrite or update a .tldr file (validates format).
tldraw_write({
path: "notes.tldr",
content: { /* tldraw file object */ }
})
tldraw_createCreate a new empty tldraw canvas.
tldraw_create({ path: "new-canvas.tldr", name: "My Canvas" })
tldraw_listList all .tldr files with page/shape counts.
tldraw_list({ recursive: true })
tldraw_searchSearch text content across all canvases.
tldraw_search({ query: "TODO", searchIn: "text" })
tldraw_get_shapesGet all shapes from a file, optionally filtered by page.
tldraw_get_shapes({ path: "notes.tldr", pageId: "page:abc123" })
tldraw_add_shapeAdd a new shape to a canvas.
tldraw_add_shape({
path: "notes.tldr",
shape: {
type: "geo",
x: 100,
y: 100,
props: {
w: 200,
h: 100,
geo: "rectangle",
color: "blue",
fill: "semi"
}
}
})
tldraw_update_shapeUpdate properties of an existing shape.
tldraw_update_shape({
path: "notes.tldr",
shapeId: "shape:abc123",
updates: { x: 200, props: { color: "red" } }
})
tldraw_delete_shapeDelete a shape from a canvas.
tldraw_delete_shape({
path: "notes.tldr",
shapeId: "shape:abc123"
})
This server works with tldraw v2 format:
{
"tldrawFileFormatVersion": 1,
"schema": {
"schemaVersion": 2,
"sequences": { ... }
},
"records": [
{ "id": "document:document", "typeName": "document", ... },
{ "id": "page:xxx", "typeName": "page", ... },
{ "id": "shape:xxx", "typeName": "shape", "type": "geo", ... }
]
}
# Install dependencies
bun install
# Run in development mode
bun run dev
# Type check
bun run typecheck
# Build for distribution
bun run build
# Run tests
bun test
./scripts/publish.sh # Auto-increment patch (0.1.0 → 0.1.1)
./scripts/publish.sh 0.2.0 # Use specific version
./scripts/publish.sh minor # Bump minor (0.1.1 → 0.2.0)
./scripts/publish.sh major # Bump major (0.2.0 → 1.0.0)
The script updates package.json, commits, tags, and pushes. GitHub Actions handles npm publish automatically via Trusted Publishers.
TLDRAW_DIRMIT © Talha Orak
Run in your terminal:
claude mcp add tldraw-mcp -- npx Yes, Tldraw MCP is free — one-click install via Unyly at no cost.
No, Tldraw runs without API keys or environment variables.
Self-hosted: the server runs locally on your machine via the install command above.
Open Tldraw on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
CSA PROJECT - FZCO © 2026 IFZA Business Park, DDP, Premises Number 31174 - 001
Security
Low riskAutomated heuristic from public metadata — not a security guarantee.