AI bug fixing · MCP
Your AI agent fixes the bugs your testers report.
FeedBug turns a visual bug report into source-level context. Its MCP server lets Claude Code, Cursor or Windsurf trace each bug to the exact file and ship the patch, all from Linear.
Free forever for solo teams. No credit card.
From bug report to pull request.
Your tester clicks the bug
One script tag on your site. A tester points at what broke and adds a comment. FeedBug captures the screenshot, console logs, failed network requests and the DOM context of the clicked element.
Your agent reads the context via MCP
The bug lands in Linear and in the FeedBug MCP server. Your AI agent pulls the CSS selector, component path, diagnostics and replay, and traces the bug to the exact file and line.
It ships the fix
The agent proposes the patch, you review, and FeedBug marks the bug resolved with the status synced back to Linear. Reporting and fixing, one loop.
Full DOM context
CSS selector path, framework component name, attributes, ancestors and siblings, captured at the moment of the report.
Works in your editor
Claude Code, Cursor, Windsurf, VS Code. One install command and your agent gets the FeedBug tools.
Diagnostics attached
Console logs, JS errors, failed network requests, device info and a 30-second session replay travel with every bug.
Resolve from the agent
Comment on the issue and mark it resolved straight from your editor, with the status synced back to Linear.
Frequently asked questions
What is MCP bug tracking?
MCP (Model Context Protocol) is the standard that lets AI coding agents call external tools. FeedBug ships an MCP server, so an agent like Claude Code or Cursor can read your bug reports, their diagnostics and the DOM context, and act on them directly from your editor instead of you copy-pasting a ticket.
How does the AI find the code to fix?
When a tester clicks a bug, FeedBug captures the HTML context of the element: its CSS selector path, the framework component name when detectable (Vue, React, Angular, Svelte), attributes, ancestors and siblings. The MCP server hands that to your agent, which maps it to the exact source file instead of guessing from a vague description.
Which AI agents does it work with?
Any MCP-compatible client: Claude Code, Cursor, Windsurf and VS Code. You add the FeedBug MCP server once with a single command, then your agent gets tools to list bugs, read their full context and diagnostics, comment, and mark them resolved.
Does AI bug fixing work with Linear?
Yes. FeedBug is built around Linear. Every bug report becomes a Linear issue with the screenshot, diagnostics and a link back to the live pin, and your agent can resolve it from the editor with the status synced back to Linear.
Is it secure to give an agent access to my bugs?
The MCP server only reads data for the project key you configure, over your authenticated FeedBug API. It exposes bug reports, diagnostics and DOM context, never your source code or secrets. The agent runs locally in your own editor.
Close the loop in two minutes.
One script tag for your testers, one MCP install for your agent. Bugs reported and fixed in the same place.