MCP feedback tool
Website feedback over MCP: let your AI read the comments
Symu 2 includes an MCP server that gives AI coding agents access to your website feedback. Claude Code, Claude Desktop, Cursor and other MCP clients can list open comments, read each one with its selector, XPath, HTML snippet, page URL and breakpoint, reply in the thread and mark it resolved.
14 days free, no credit card required.
What the Symu 2 MCP server does
MCP (Model Context Protocol) is an open standard for connecting AI assistants to external tools. The Symu 2 MCP server exposes your website feedback as tools, so an AI coding agent can read open comments with full element context, reply to them and resolve them.
You stay in control of the code. The agent asks Symu 2 what needs fixing, makes the change in your repository with its usual tools, and reports back on the comment thread where your client can see it.
Connect Claude Code in one command
- Open “MCP for AI” in the header of the Symu 2 panel.
- Copy the command and run it in your terminal. It has this form:
$ claude mcp add --transport http symu https://symu.co/mcp/YOUR_TOKEN - Ask the agent, for example: “List the open Symu comments for the Acme website and fix them one by one. Reply on each comment with what you changed and resolve it.”
Claude Desktop uses the same link as a custom connector. The token is personal: you can create named tokens and revoke them under Settings, AI access.
Connect Cursor and other MCP clients
Any MCP client that supports remote servers over HTTP can use the same link. In Cursor, add it to your mcp.json:
{
"mcpServers": {
"symu": { "url": "https://symu.co/mcp/YOUR_TOKEN" }
}
}
Client configuration formats change between releases, so check your client’s documentation for the current syntax. If a tool has no MCP support, use copy as AI prompt on a single comment or export open comments as markdown and paste the result.
Tools available to the agent
| Tool | What it does |
|---|---|
list_projects | Lists your projects and their screens (websites or designs) with open and resolved comment counts. The agent calls it first to find the right ids. |
list_comments | Lists comment threads, open ones by default, filterable by page URL, breakpoint (desktop, tablet, mobile) and date. |
get_comment | Returns one comment with all replies, its element locators (selector, XPath, HTML snippet), page URL, breakpoint, screenshot link and a ready-to-use prompt. |
export_open_comments | Returns all open comments of a project or screen as one markdown task list, grouped by page URL, with element, selector and breakpoint. |
reply_comment | Posts a reply in the comment thread, marked as written via the agent. |
resolve_comment | Marks the thread as resolved when the fix is done, or reopens it. |
get_screen | Returns one screen: type, website URL, cover screenshot, comment counts and the pages that have comments. |
What an agent sees for one comment
#12 [open] by Anna (client) · Website "Acme" › /pricing Comment: "Make this button green and say 'Start free'" Page: https://acme.com/pricing · desktop 1440×900 Element: <button.btn-primary#signup> "Sign up" Selector: #plans > div.card:nth-of-type(2) > button
The format above is illustrative. The point is the context: the agent can search the repository for the selector, the class or the visible text and edit the right file.
Where this fits
New to Symu 2? Start with the website feedback tool overview. Building with Lovable, Bolt, v0 or Cursor? Read the vibe coding workflow. For a step-by-step walkthrough see how to give your AI coding agent feedback on a live website.
Other feedback tools document MCP support too, including Pastel, BugHerd and Marker.io. We compare them honestly on the alternatives pages.
What is an MCP server for website feedback?
The Model Context Protocol (MCP) is an open standard that lets AI assistants call tools in other applications. An MCP server for website feedback exposes your review comments as tools, so an AI coding agent can fetch them, understand which element each one refers to and report back.
How do I connect Symu 2 to Claude Code?
Open “MCP for AI” in the Symu 2 panel header, copy the one-line command and run it in your terminal. It has the form claude mcp add --transport http symu followed by your personal MCP link. Claude Desktop and other clients use the same link.
Which tools does the Symu 2 MCP server provide?
list_projects, list_comments, get_comment, export_open_comments, reply_comment, resolve_comment and get_screen. Together they let an agent browse your projects, read open feedback with full element context, answer in the comment thread and close the comment after the fix.
Can the AI see which element a comment refers to?
Yes. Each comment carries the CSS selector, XPath, an HTML snippet of the element, the page URL and the breakpoint (desktop, tablet or mobile) where it was placed. That is usually enough for an agent to locate the code responsible.
Is my feedback safe when I connect an AI agent?
The MCP link is a personal access token that you can create and revoke in the panel under Settings, AI access, and the agent only sees the projects that your own account can see. Revoke the token and the access stops immediately.
Does Symu 2 work with Cursor, Codex or Windsurf?
Any MCP client that supports remote servers over HTTP can use the same link. If a tool does not support MCP, copy a comment as an AI prompt or export all open comments as a markdown task list and paste it in.
See how Symu 2 works on your own website.
Paste a URL, invite a client and let your AI read the feedback. 14 days free, then $19/month.