Features

AI access

Hand feedback to AI: copy it as text, connect Claude by URL, or connect Claude Code, Codex or Cursor over MCP.

Feedback is only useful once it is fixed. AI access is the set of ways to get a project's feedback out of layernote and into whatever works on it: a chat you paste into, Claude reading your files and comments, or a coding agent working in your repository.

In a project, click the AI icon in the top bar, or choose Hand off to AI in the project menu. Each row in that panel works on its own. Settings, AI access holds the same connections for your whole account.

Ways to hand off feedback

	Setup	Best for
Copy for any AI	None	Pasting the open feedback into any AI chat, a pull request or a ticket
Connect Claude	A URL, then sign in	Claude on the web or in the desktop app: summaries, change lists, replies, a review of a file
Coding agent	A token and one command	Claude Code, Codex or Cursor fixing feedback in your codebase

layernote's MCP server only supplies data and actions over the Model Context Protocol. It never calls a model itself, so it works with whichever AI you already use and adds nothing to your AI bill.

Copy for any AI

Click Copy. layernote puts every open item on your clipboard as plain text: where it is and the comment thread. Paste it into any AI chat, a pull request, or a ticket.

Connect Claude

Claude connects to layernote by URL. There is no token to copy and nothing to install.

Copy the connector URL
Open Settings, AI access and copy the Connector URL, or click Copy URL on the Claude row of Hand off to AI. The URL is:
```
https://app.layernote.io/mcp
```
Add it in Claude
In Claude, open Settings, Connectors, click Add custom connector, name it layernote and paste the URL. On a Claude Team or Enterprise plan, your Claude admin may have to add the connector for the organization first.
Sign in and allow
Claude sends you to layernote. Sign in, pick the workspace and click Allow. Claude can then read the comments, files and screenshots in the projects you reach in that workspace, and add comments and replies, change statuses and assign comments. It can also set a review up for you: make a project, bring in screens from Figma and give you the share link for your client. It does all of that under your name, and the comments it writes are marked "via AI".

Every app you connect this way is listed under Settings, AI access, with a Disconnect button.

The file viewer

On a Static files project, ask Claude to show you a file and it opens in a viewer inside the chat: the picture with its open comments as pins, a PDF page by page, a video moment by moment, and a switch between versions of the file. A Figma project's frames open the same way. Click a pin to read the comment, resolve it, or ask Claude about it. Claude on the web and the desktop app show the viewer; in Claude Code the same request gives you text.

Claude desktop extension

Would you rather use a token in the Claude desktop app? Create a token under Settings, AI access and click Download extension in the dialog that shows it. Open the file, click Install, and paste your token when Claude asks for it. You need no Node.js and no config file.

Claude Code, Codex and Cursor

A coding agent reads the feedback, changes the code in your repository and reports back in the thread.

Claude Code can connect by URL, like Claude itself. Run this, then run /mcp in Claude Code to sign in:

claude mcp add --transport http layernote https://app.layernote.io/mcp

Codex, Cursor and other tools use a token.

Create a token
Open Settings, AI access, click Create token, pick the workspace and give the token a name, for example the machine it will live on. In a project, Connect on the Coding agent row of Hand off to AI makes the same kind of token for the project's workspace.
The token starts with mcp_ and is shown once, so copy it right away. It is yours, for one workspace: the agent reaches every project you reach there, and everything it does is in your name. A project that is not in a workspace gets a token for that project only.
Register the server with your agent
The token dialog gives you a ready-made command for Claude Code. For other tools use the matching snippet below, replacing mcp_xxx with your token.
claude mcp add layernote --env LAYERNOTE_TOKEN=mcp_xxx -- npx -y @layernote/mcp
codex mcp add layernote --env LAYERNOTE_TOKEN=mcp_xxx -- npx -y @layernote/mcp
{ "mcpServers": { "layernote": { "command": "npx", "args": ["-y", "@layernote/mcp"], "env": { "LAYERNOTE_TOKEN": "mcp_xxx" } } } }
{ "mcpServers": { "layernote": { "command": "npx", "args": ["-y", "@layernote/mcp"], "env": { "LAYERNOTE_TOKEN": "mcp_xxx" } } } }
For Cursor, save the JSON as .cursor/mcp.json in your repository. The server talks to https://app.layernote.io by default; set LAYERNOTE_API_URL as well if you run layernote somewhere else.
In the Claude desktop app the extension is the easy way. To add the server by hand instead, open Settings, Developer, Edit Config and paste the JSON into claude_desktop_config.json. If the file already has an mcpServers block, add layernote inside it. Then quit Claude completely and open it again. The app starts the server with Node.js, so Node.js 18 or later has to be installed on that computer.
Claude can't start the server
If the desktop app says the layernote server failed, with spawn npx ENOENT or env: node: No such file or directory, it can't find Node.js: an app on a Mac doesn't see the folders your Terminal knows. Run which npx in Terminal, use that path as command, and give the server its folder as PATH, as below. Replace /opt/homebrew/bin with the folder which npx showed. Or use the extension, which needs no Node.js.
```
"command": "/opt/homebrew/bin/npx",
"args": ["-y", "@layernote/mcp"],
"env": {
  "LAYERNOTE_TOKEN": "mcp_xxx",
  "PATH": "/opt/homebrew/bin:/usr/bin:/bin"
}
```

What to ask

Inside your own repository, tell the agent something like: pull the open feedback from my layernote project and fix it. It lists the items, opens each one with its screenshot and selector, changes the code, and reports back in the thread. When the connection reaches more than one project, name the project in your request.

There doesn't have to be code to change. Ask Claude things like:

Summarize the open feedback in my layernote project.
Which mockups still have open points, and what matters most per file?
Turn the comments on the launch video into a checklist for the editor.
Did version 2 of the homepage take care of the comments on version 1?
Review this PDF and pin your comments on it.
Review the checkout screens in my Figma project and pin what you'd change.
Make a project for the new checkout, bring in its screens from this Figma link and give me the link for the client.

Apps that show ready-made prompts list four of them: a summary of the open feedback, a change list for the video editor, a change list for the designer, and a weekly overview of what was resolved, what is new and who is waiting on whom.

Available tools

When a connection reaches more than one project, every tool but list_projects and create_project takes the project's name or id.

Tool	What it does
`list_projects`	The projects the connection reaches, with their type and how many open comments each has.
`create_project`	Makes a project in your workspace: a Figma design, Static files, or a website with its address. It counts against your plan like a project made in the app. It needs the URL connection or a workspace token: a token for a single project can't make one.
`list_feedback`	Lists comments, newest first: by default the open ones (Open, In Progress, In Review), with author, assignee, due date, thread and where each one is. Filters by status, file, author, assignee and date.
`get_feedback`	One comment with its full thread, its screenshot and the pictures attached to it, returned as images the agent can look at.
`list_files`	The files of a Static files project (images, PDFs and videos) in the order of the Files sidebar, with how many open comments each has. A file with versions is listed once, at its latest version. On a Figma project: its frames, in the order of the Frames sidebar.
`get_file`	A file or a Figma frame with its open comments as numbered pins: an image or frame whole, a PDF page by page, a video as the frames its comments are on.
`compare_versions`	An older version's open comments, each beside the same spot in a newer version, to tell which ones the new version took care of.
`show_file`	Opens a file or a Figma frame in the viewer inside the chat, with pins to click.
`add_comments`	Places comments on a file or a Figma frame as pins, on a spot of a picture, a PDF page or a moment in a video. The team gets one notification for the whole batch.
`upload_file`	Adds an image or PDF to a Static files project, as a new file or as the next version of one: from links layernote fetches (up to 10 at once), or through an upload panel in the chat (up to 8 MB a file). Images in JPEG, PNG, WebP, GIF or AVIF, and PDFs; videos go up in the app.
`add_attachment`	Attaches an image or PDF to a comment.
`list_figma_frames`	The screens in a Figma file, with their size and whether the project already has them, or Figma changed them since.
`import_figma_frames`	Brings screens from a Figma file into a Figma project, as Add your Figma design does in the app. A screen already there is brought up to date when it changed in Figma, and keeps its comments.
`share_project`	The project's share link for your client: anyone with it can comment without an account, or only look. A link your team paused stays closed unless you ask Claude to open it again.
`add_feedback_comment`	Posts a reply in a comment's thread, for example what was changed or why something is blocked.
`update_feedback`	Assigns a comment to someone on the project, or sets its priority or due date.
`list_people`	Who can work on the project's feedback, and so who can be assigned.
`set_feedback_status`	Moves a comment to any column on the project board, custom columns included.
`resolve_feedback`	Marks a comment resolved.
`wont_fix_feedback`	Closes a comment as won't fix with a mandatory reason, for false positives and out-of-scope requests.

The tools map onto the same statuses your team sees on the board, so AI's work shows up like anyone else's: replies in the thread, cards moving across columns, and your name on everything it does.

Importing from Figma uses your own Figma connection in layernote, not a Figma connector in the chat. Connect it once: open a Figma project, click Add your Figma design and sign in to Figma. Until then, Claude tells you to do that step yourself.

Access and security

A token or a connected app belongs to one person and one workspace. It reaches only the projects that person reaches there, and everything it does carries their name. Leaving the workspace stops it.
Tokens are stored hashed. layernote can never show a token again, only revoke it. Revoke a token or disconnect an app under Settings, AI access, and it loses access at once.
Workspace owners and admins see every token and connected app of the workspace's people under Settings, AI access, Your team, and can revoke any of them. An admin can't revoke the owner's. Revoking doesn't keep someone out for good: removing them from the workspace does.
AI access works while the workspace's plan is active.
The MCP server you run with a token runs on your machine and only talks to layernote. It never sends your code anywhere. A connection by URL runs on layernote's side, so nothing is installed.