Connect OmniCanvas to AI assistants like Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. Available on the Power plan.

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants securely access your tools and data. With OmniCanvas's MCP server, your AI assistant can read, create, search, and organize your notes — all with your explicit permission on every action.

Setup

Open OmniCanvas and go to Settings → API Access
Click Create to generate an API key
Copy the MCP configuration that appears (it includes your key)
Paste it into your AI client's MCP settings file

Or configure it manually — add this to your client's MCP config:

MCP Configuration

{
  "mcpServers": {
    "omnicanvas": {
      "url": "https://app.omnicanvasnotes.com/api/v1/mcp",
      "headers": {
        "Authorization": "Bearer pk_live_your_key_here"
      }
    }
  }
}

Client-Specific Setup

Claude Desktop

Open Settings → Developer → Edit Config and paste the configuration above into your claude_desktop_config.json. Restart Claude Desktop.

Claude Code (CLI)

Add to your project's .mcp.json or global ~/.claude/settings.json under the mcpServers key.

Cursor

Go to Settings → MCP and add a new server. Use the URL and Authorization header from the config above.

Other Clients

Any client that supports the MCP 2025-03-26 spec with Streamable HTTP transport will work. Point it at https://app.omnicanvasnotes.com/api/v1/mcp with your API key as a Bearer token.

Available Tools

Your AI assistant can use 14 tools to interact with your notes. Here's what each one does:

list_notes

List your notes with optional filtering by tag, folder, or search query.

Parameter	Type	Description
tag	string	Filter by tag
folderId	string	Filter by folder
search	string	Search titles and OCR text
limit	integer	Max results (1–100, default 50)
offset	integer	Skip N results

get_note

Get a note by ID with full content including canvas data, tags, and folders.

Parameter	Type	Description
id*	string	Note ID
includeCanvas	boolean	Include canvas data (default true)

create_note

Create a new note with optional title, tags, folders, and canvas data.

Parameter	Type	Description
title	string	Note title
tags	string[]	Tags to apply
folderIds	string[]	Folders to add the note to
canvasData	object	Excalidraw canvas JSON

update_note

Update a note. Only provided fields are changed. Tags and folders replace existing values.

Parameter	Type	Description
id*	string	Note ID
title	string	New title
tags	string[]	New tags (replaces all)
folderIds	string[]	New folders (replaces all)
isPinned	boolean	Pin or unpin

delete_note

Soft delete a note (moves to trash). Pass permanent=true to permanently delete.

Parameter	Type	Description
id*	string	Note ID
permanent	boolean	Permanently delete (default false)

list_folders

List all your folders with note counts.

create_folder

Create a new folder, optionally nested inside another.

Parameter	Type	Description
name*	string	Folder name
parentId	string	Parent folder for nesting

search_notes

Full-text search across note titles and OCR text. Returns excerpts with matching context.

Parameter	Type	Description
q*	string	Search query (min 2 chars)
limit	integer	Max results (default 20)

list_tags

List all tags across your notes, sorted by usage frequency.

export_note

Export a note as full Excalidraw JSON without size limits.

Parameter	Type	Description
id*	string	Note ID

list_versions

List version history for a note (12-month retention on Power plan).

Parameter	Type	Description
noteId*	string	Note ID

restore_version

Restore a note to a previous version. Current state is auto-saved first.

Parameter	Type	Description
noteId*	string	Note ID
versionId*	string	Version ID to restore

clip_url

Clip a web page or YouTube video. Extracts title, content, and optional transcript. YouTube URLs are auto-detected.

Parameter	Type	Description
url*	string	URL to clip
mode	string	"visual" (screenshot + text) or "text" (default)
includeTranscript	boolean	Fetch YouTube transcript (quota-limited)

youtube_transcript

Fetch the transcript for a YouTube video. Quota-limited per month.

Parameter	Type	Description
videoId*	string	YouTube video ID (11 characters)

Resources

MCP resources let your AI assistant read structured data from your account:

note://{id} — full note content with metadata
folder://{id} — folder details with its note list
tags://all — all your tags with usage counts

Example Use Cases

Once connected, you can ask your AI assistant things like:

"Search my notes for anything about project Alpha"
"Create a new note with today's meeting summary and tag it #meetings"
"Clip this article and save the key points to my Research folder"
"What did I write about database migrations last week?"
"Get the transcript from this YouTube video and add it to my notes"
"Show me all my notes tagged #ideas and summarize the common themes"
"Restore my project plan to yesterday's version"

Limits

Rate limit: 100 requests per minute
Canvas read size: Notes larger than 50KB return truncated canvas data — use export_note for the full file
Encrypted notes: Canvas data is returned as null (encryption/decryption happens client-side)
YouTube transcripts: 200 per month on Power plan

Security

Your API key is hashed (SHA-256) before storage — we never store the raw key. Each request is authenticated and scoped to your account only. You can revoke keys instantly in Settings. The MCP server uses Streamable HTTP transport over HTTPS.

MCP Server