Work through client feedback
with your AI assistant
Connect Feedbucket to Claude, ChatGPT or Cursor. Your AI gets the screenshot and technical details of every piece of client feedback, so it can help you fix issues and reply to clients.
MCP is included in the 14-day free trial and every paid plan. No credit card needed.
Feedbucketnow
Re: The menu doesn't open when I tap it on my phone.
Alex: Fixed. Could you check the menu on your phone again?
ResolvedTo jane@acme.com
Your AI gets the same context your developers do
Every feedback item comes with an annotated screenshot, the page URL, device details and console logs. Your AI reads all of it, so nobody has to ask the client to explain again.
Fix client feedback without leaving your editor
Ask Cursor, Claude Code or Codex to pick up a feedback item. It sees the annotated screenshot, the element the client clicked, the viewport size and any console errors. With that, it can find the right file in your repo and propose a fix for you to review.
Screenshots come through as images your AI can look at
The CSS selector points it to the exact element
Console errors from the moment the client submitted come along too
Try asking
Look at the newest open feedback on the Acme project and fix it in this repo.
- feedback-get
- feedback-get-console-log
Triage a whole review round in one message
Review rounds get messy. Three people from the client report the same broken image, and the urgent bug sits under a pile of copy tweaks. Ask your AI to go through the open feedback, find the duplicates and sort the rest by page, type or urgency. It can look at the screenshots too, so it catches two reports of the same problem written in different words.
Spot duplicates, even when clients describe them differently
Tag each item from your project's tag list, like Bug, Design or Copy
Ask what changed since a given date, including new comments
Try asking
Go through the open feedback on Acme. Flag duplicates, tag each item as Bug, Design or Copy, and list the bugs first.
- feedback-list
- feedback-get
- feedback-update
Reply to clients and resolve feedback from the chat
When a fix is ready, ask your AI to reply in the feedback thread, attach a screenshot of the result and resolve the item. It works like any other comment in Feedbucket. The client sees the reply on the website and gets an email notification. The comment and the new status also sync to Asana, ClickUp, Jira or whichever tool you've connected.
Attach screenshots, PDFs or other files to a reply
Add members-only notes that clients never see
Comments and status changes sync to your project management tool
Try asking
Tell Jane the menu is fixed, attach this screenshot and resolve her feedback.
- attachment-create
- comment-create
- feedback-update
Set up client projects by asking
Starting a new site? Ask your AI to create the project, set the widget language and add your usual tags. If the staging site is password protected, it can add the basic auth details too. When the site launches, ask it to hide the widget or archive the project.
Change widget visibility, placement, language and onboarding text
Restore an archived project when the client comes back for a redesign
Try asking
Create a Feedbucket project for staging.nordicbakery.se, set the widget to Swedish and add the tags Bug, Design and Copy.
- projects-create
- project-update
- project-archive
"Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua."
Connect your AI in three steps
Each person on your team connects once, in the AI app they already use.
01. Add Feedbucket to your AI app
Open MCP settings in your Feedbucket account and choose your app. Cursor and VS Code install with one click. For Claude, ChatGPT and Codex, paste the server URL.
02. Approve the tools you allow
Sign in to Feedbucket and choose which tools your AI can use. Let it read feedback freely, and have it ask before it comments, resolves or changes a project.
03. Start asking
Open a new chat and ask about your projects. A good first question: "Show me the open feedback on my latest project."
Works with
- Cursor
- VS Code
- Claude Code
- Claude Desktop
- Claude.ai
- ChatGPT
- Codex
Using another app? Any MCP client that supports remote servers can connect with the server URL. See the setup guide for each app
Frequently asked questions
MCP (Model Context Protocol) is an open standard that lets AI assistants use tools from other apps. Once you connect Feedbucket, your AI assistant can read your feedback and act on it directly. You stop copying screenshots and comments into the chat by hand.
Feedbucket is a website feedback tool for web design agencies. Clients leave feedback directly on the website as annotated screenshots or screen recordings, without an account or browser extension. Each item includes the page URL, browser, device and console logs, and syncs both ways with Asana, ClickUp, Jira and other project management tools. See how Feedbucket works.
Cursor, VS Code, Claude Code, Claude Desktop, Claude.ai, ChatGPT and Codex all have step-by-step guides in our documentation. Any other app that supports remote MCP servers can connect with the server URL https://dashboard.feedbucket.app/mcp.
Yes. The MCP is included in every paid Feedbucket plan at no extra cost. Your AI app's own subscription or usage fees still apply. See pricing for what each plan includes.
Yes. The MCP works from day one of the 14-day free trial, and you can start without a credit card. Start your free trial, then open Account > MCP in the dashboard to connect your AI.
No. Clients leave feedback on the website the same way they always have. Only the people on your team who want to use the MCP connect an AI app.
Your AI works with your permissions on your current Feedbucket team. It can list, create and update projects, read feedback with screenshots and console logs, write, edit or delete comments, and resolve or reopen feedback. It has no tool for deleting feedback or projects. Projects can only be archived, and archived projects can be restored. The full list of tools is in the documentation.
The MCP works on your Feedbucket data. Settings your AI changes, like widget visibility or language, apply to the Feedbucket widget on that site. It has no access to your website's code. If you use a coding agent like Cursor or Claude Code that has access to your repository, it can propose code changes, and you review and deploy them through your usual workflow.
Your AI app reads the feedback it needs to answer your request, the same as if you pasted it into the chat yourself. How that data is stored and used follows your AI provider's terms and your settings there.
Each person connects their own AI app with their own Feedbucket account. Comments and changes show up under the name of the person who made them, so your team and clients can see who replied.
Go to Account > Access tokens in the Feedbucket dashboard. You'll see every token and every app connected through sign-in, and you can revoke any of them. Access tokens also expire automatically after one year.
The Feedbucket MCP documentation lists every tool, what it does, and the setup steps for each AI app.
Give your AI the full picture.
Start your 14-day trial with the MCP included. No credit card needed.
