The release is on Friday. The tester found a bug on Wednesday. The ticket says "checkout is broken." Everything after that sentence is a conversation: which browser, which account, what did the network tab say. Release testing, sprint testing and UAT all fail the same way, and this page covers all three.
The gap: a tester sees it, the fixer can't
Your tester knows what the page should do. They don't know what the console said or which request failed. Your developer, or your coding agent, needs exactly that. Without it the ticket bounces back with "cannot reproduce" and the release window shrinks.
How it works
- Add the widget to staging. One script tag on the environment your testers use.
- Testers click, describe and (on paid plans) mark up. No DevTools, no account for the reporter.
- The ticket arrives with evidence. The element, console errors, requests with bodies, the steps before the bug and the browser and build. On paid plans, also the screenshot.
What the ticket contains
- The element the tester pointed at, with errors tied to it
- Console and JavaScript errors with file and line
- Requests with status, URL and bodies, failed and successful
- On paid plans: an annotated screenshot of what the tester saw
- Session replay of the steps leading up to the report
- URL, browser, OS, viewport
Sprint testing, release testing and UAT
Sprint testing: QA files reports against the build that is on staging. The developer opens the ticket and starts from the evidence rather than a chat thread.
Release testing: the last pass before go-live is time-boxed. Reports that arrive complete let you decide fix, defer or hold in minutes.
UAT: business testers know what the system should do but not how to write a bug report. They circle the wrong number and describe what they expected; the technical context is captured for them.
Let your agent take the fix
If you use Cursor, Claude Code or Codex, connect the SnagRelay MCP server once. Then tell the agent to fix a report by its tracker key. It reads the same evidence a human would, makes the change, and marks the report resolved. The tester gets a "fixed" email, so they can retest.