API and MCP

LetDraw’s REST API and MCP server let external tools and AI clients (Claude, Cursor) work with your diagrams: list, read, create, update, generate a diagram from code, convert back to code, share, and generate from a prompt.

Who can use it

The API and MCP are available to Pro and Enterprise users. On self-hosted installations, every signed-in member can use them. Free and guest users cannot create tokens.

1. Create a token

Sign in to LetDraw with a Pro/Enterprise account.
Go to Settings -> Developer.
Create a token: give it a name, pick the scopes (see below), and optionally restrict it to a single workspace.
The generated ld_live_... token is shown only once. Save it somewhere safe. If it leaks, revoke it from the same page.

Scopes

Scope	Permission
`diagrams:read`	Read workspaces, folders, and diagrams
`diagrams:write`	Create, update, delete, and share diagrams
`export`	Export a diagram to Mermaid/D2 code
`ai:generate`	Generate a diagram from a prompt (AI)

2. Authentication

Send the token as a Bearer value in the Authorization header on every request:

Authorization: Bearer ld_live_...

No extra apikey header is needed.

REST API

Base URL: https://api.letdraw.com/api-v1

All responses are JSON. Errors return with a standard body: { "error": { "code", "message" } }.

Method	Path	Description	Scope
GET	`/`	API info (scopes, endpoints)	any
GET	`/workspaces`	Your workspaces (with role)	read
GET	`/folders?workspace=<id>`	Folders in a workspace	read
GET	`/documents?workspace=<id>`	Diagrams in a workspace (paginated)	read
GET	`/documents?folder=<id>`	Diagrams in a folder (paginated)	read
POST	`/documents`	Create a diagram	write
GET	`/documents/:id`	Diagram + scene (elements)	read
PATCH	`/documents/:id`	Update name and/or elements	write
DELETE	`/documents/:id`	Delete a diagram	write
POST	`/documents/from-code`	Generate a diagram from code	write
POST	`/documents/generate`	Generate a diagram from a prompt (AI)	ai:generate
GET	`/documents/:id/export?format=mermaid\|d2`	Export a diagram to code	export
POST	`/documents/:id/share`	Create a public read-only link	write
DELETE	`/documents/:id/share`	Remove the public link	write

Pagination

List endpoints take ?limit= (1-100, default 50) and ?cursor=. The response returns a nextCursor; pass it as cursor to get the next page.

Rate limit

120 requests per minute per token. Exceeding it returns 429 rate_limited.

Examples

# List workspaces
curl -H "Authorization: Bearer ld_live_..." \
  https://api.letdraw.com/api-v1/workspaces

# Generate a diagram from a docker-compose
curl -X POST -H "Authorization: Bearer ld_live_..." -H "Content-Type: application/json" \
  -d '{"workspaceId":"<id>","code":"services:\n  web:\n    image: nginx\n    depends_on: [db]\n  db:\n    image: postgres\n"}' \
  https://api.letdraw.com/api-v1/documents/from-code

# Export a diagram to Mermaid
curl -H "Authorization: Bearer ld_live_..." \
  "https://api.letdraw.com/api-v1/documents/<id>/export?format=mermaid"

MCP server

Model Context Protocol (MCP) clients (Claude, Cursor) can connect to LetDraw and work with your diagrams through tools.

URL: https://api.letdraw.com/mcp
Transport: Streamable HTTP (stateless JSON-RPC 2.0)
Auth: Authorization: Bearer ld_live_...

Tools

Tool	Description
`list_workspaces`	Accessible workspaces
`list_diagrams`	Diagrams in a workspace/folder
`get_diagram`	Diagram + scene
`create_diagram`	Create a diagram from an elements array (LetDraw or Excalidraw format, auto-detected)
`update_diagram`	Update name/elements
`delete_diagram`	Delete a diagram
`diagram_from_code`	compose, kubernetes, dot, plantuml, terraform, helm, sql -> diagram
`export_to_code`	Diagram -> Mermaid/D2
`share_diagram`	Create a public link
`unshare_diagram`	Remove the public link
`generate_from_prompt`	Diagram from natural language (AI)

Client setup

You can copy ready-made config blocks with your own token from the Settings -> Developer page. Replace ld_live_... below with your own token.

Cursor

~/.cursor/mcp.json (global) or .cursor/mcp.json in the project root:

{
  "mcpServers": {
    "letdraw": {
      "url": "https://api.letdraw.com/mcp",
      "headers": { "Authorization": "Bearer ld_live_..." }
    }
  }
}

Claude Desktop

Claude Desktop connects to the remote HTTP server through the mcp-remote bridge (Node.js required). Config file:

Windows: %APPDATA%\Claude\claude_desktop_config.json
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "letdraw": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://api.letdraw.com/mcp",
        "--header", "Authorization:Bearer ld_live_..."
      ]
    }
  }
}

Restart the client after saving. Example prompts:

“List my LetDraw workspaces.”
“Turn this docker-compose into a diagram in letdraw: …”
“Export diagram as Mermaid.”
“Generate an OAuth login flow diagram.”
“Draw this cloud architecture in letdraw: …” (free-form composition)

Telling the AI to “draw it in LetDraw”

create_diagram and update_diagram accept the elements array in two formats: LetDraw’s own element schema or the Excalidraw format. The format is auto-detected, and Excalidraw elements are converted to LetDraw elements on the server side (including colored nested containers, shape-bound labels, arrows with bindings, text, and frames).

What this means: when you tell ChatGPT, Gemini, or Claude to “draw this architecture in LetDraw”, the model produces the Excalidraw JSON it already knows very well from its training data and sends it over MCP; LetDraw draws the exact equivalent. Whatever the model’s layout quality is, that is what you get in LetDraw.

Embedding a diagram (render.svg)

You can embed a publicly shared diagram as SVG in places like a README or Notion. First get a share token with share_diagram (or REST .../share), then:

https://api.letdraw.com/render-svg?token=<share-token>

This endpoint returns image/svg+xml and requires no API token (only shared diagrams are rendered), so you can use it directly inside <img src="...">.