Qure: Desktop App for Generating E2E Tests from Recorded Browser Flows
Qure is a desktop application built at JetBrains that's currently in closed beta. It generates end-to-end web test code from recordings made in its built-in browser.
How It Works
Instead of describing test flows in text for AI agents, you record your manual QA scenarios by interacting with your product in Qure's built-in browser. The AI then turns that recording into test code.
Key differentiators from typical AI test generation:
- No prompt engineering required
- No MCP setup needed
- No explaining your repository structure in chat - you point it at your project and go
- The AI matches recordings against your existing codebase to find page objects, helpers, and constants
- When tests fail, the agent reads real failure output and fixes issues with actual error and application context
Additional Features
Beyond recording, you can also:
- Refactor existing tests
- Update tests
- Write new tests from descriptions
Technical Details
The tool is web-only and works best with Playwright. According to the developers, if your project only has a few dozen tests, Claude Code might be sufficient. Qure makes more of a difference on larger codebases with existing test infrastructure.
The developers note this is an experimental product in closed beta and are open to questions about where the approach breaks or has limitations.
📖 Read the full source: HN AI Agents
👀 See Also
ClawedBack: OpenClaw Port Running Inside Claude Code
ClawedBack is a clean-room port of OpenClaw that runs inside Claude Code, providing first-party prompt caching and rate limits. It matches 19 out of 23 of OpenClaw's built-in tools and is fully ClawHub compatible with mandatory security scans for imports.
/goal for Claude Code: persistent tasks with adversarial review
A /goal command for Claude Code that keeps it working on a long task across many turns, with an optional separate Claude session reviewing the final result to prevent false completion.
MCP Server Adds Persistent Memory with Retrieval Scoring to Claude Code
A developer built an MCP server called engram-mcp that gives Claude Code persistent memory across sessions and projects, featuring automatic retrieval scoring based on outcome success and drift detection for stale knowledge.
Creation OS: A Local σ-Gated LLM Runtime That Lets Models Say ‘I Don’t Know’ Instead of Hallucinating
Creation OS wraps local LLMs (BitNet, Qwen, Gemma, any GGUF) with a σ-gate that measures multiple uncertainty channels and decides ACCEPT, RETHINK, or ABSTAIN per output. No cloud, no API. TruthfulQA accuracy improved ~29% via selective regeneration.