Introducing Expect Test your agent code in a real browser Works with your agent. Fully open source

Agents are bottlenecked because it needs YOU to test its changes Expect automates that: → Watch and steer on a live browser (with cookies!) → Performance, accessibility, design & security → Works inside Claude Code, Codex, etc.

Install it via your terminal: npx expect-cli@latest init

we're open source + love feedback. reply if you hit an issue and I'll get it resolved live

a lot of fun designing this!

cooked

porting this to Same

lfg

this demo is real btw (it's literally a live agent cursor on the screen)

Expected

This is really cool!! Is there any plans to allow it to help with writing tests? I use Pest PHP (which uses Playwright under the hood for browser testing), I’d love to be able to click around and have it convert that to a test for complex functionality. Unless I’ve mistaken the testing part of Expect

would you be willing to share more details via DM? happy to see if we can make it work

This is so useful! Can't wait to try it

lmk how it goes! should be super easy to install + try!

rabbit r1 LAM for testing but actually good

bro RABBIT?

it was destined to be fr

unExpectedly really good

wow that is so funny Isabelle

@samedotdev Used Expect to do the entire QnA of an app that I’ve migrated to a new framework. I didn’t look at anything, worked like a charm. Expect + Dev Agent was amazing

background agent -> expect -> video with changes to slack

lmk if you need help setting up this flow!

wohhhhhh , thats so amazing

how much of perf/a11y/design/security checks are heuristics vs real audits? can i weight/disable some?

it's agent based on runtime info. agent has special tools (like profiler, axe core) and the agent audits

clean idea!

cool!

This is good!!! Gonna go try this out rn!

Cooking

Cool, how is this different from D3k?

Is it just a skill? How is it different than puppeteer?

dope combo playwright mcp + sync' with coding agents run logs!

Dumb question: how’s it different from me telling CC to write the QA checklist and have it self-verify by using playwright or agent-browser

What's the biggest gap you found between simulated and real browser behavior that pushed you to build this?

going to try this tomorrow, it’s been on my list for a few days now

How do u make a living @aidenybai ? You’re open sourcing your tools

rad. TIL about million co. i dont really buy into the whole 'mcp bad' thing. but, any particular reason this is very MCP driven? at least, according to the skill file it seems that way

not sure why this got so little views, looks amazing, going to give it a spin!

What do you use for video editing, the vids are fire!!

I learned this the hard way. Browser testing doesn't solve agent reliability, it just shifts the fragility.

Can I run expect for something like a Salesforce project? I often have to test a lot of LWCs which is a big pain in the ass.