← Complete register experiment · Past project
SaaS Maker directory
Browser Agent Testing
A completed local web and iOS experiment comparing browser-agent strategies with independent verification.
Maker note
Why I made this.
I ran Browser Agent Testing to compare automation by verified feedback time, not by how quickly a driver can finish clicks.
Product contract
What this product promises.
Browser Agent Testing compares local web and native UI automation with reproducible, product-owned verification.
- For
- Coding agents and maintainers comparing local browser and native UI testing tools.
- Outcome
- Choose faster and cheaper test strategies without silently losing functional or visual defect detection.
- How
- Bounded manifests and receipts run saved or adaptive workflows, then independent product verifiers check UI, persistence, process and visual outcomes.
- Proof
- The completed experiment publishes a sourced 79-tool catalogue, 26 measured journey arms with 118 verified completions across 129 attempts, seven additional probes, ten retained experiment records, exact observed pins and a live agent-first guide. The measured evidence retains Playwright and Maestro as baselines.
- Next action
- Explore the finished experiment or clone the public repository; replay a documented arm only when a concrete product change justifies it.
Public anatomy
What it is made of.
- Form
- Completed benchmark and static guide
- Platforms
- Web · iOS · CLI
- Prominent tools
- Node.js · Cloudflare Workers · Playwright · Maestro
- Deployment
- Live archive Cloudflare Workers
- First retained commit
- Latest retained commit
Maintained evidence
Follow the work.
Public destinations and repository evidence, when this project publishes them.
Repository publication
Read the issues.
Public issues become a browsable, read-only record here. Open the source issue to comment, react, or follow the work.
Every fact here comes from the project’s reviewed public record. Private work and operational controls stay private.