Jakub Stonavsky

Commented on VirtuProbe Studio · Aug 2, 2026

FYI if you want to see VP work with an agent to identify why a test failed (in our new FooBarHats test lab), check out this clip: https://dl.virtuprobe.studio/promo/agent_investigates_discount_error.mp4

Commented on VirtuProbe Studio · Aug 2, 2026

@zehan - I just sent you an email ... suggestions and feedback are 100% appreciated!

Commented on findalternate · Jul 29, 2026

Hey! Looks like a really nice product (with a very fresh UI compared to the competition :-)). What is your plan regarding seeding and maintaining the database? Do you plan going fully AI or is there going to be some level of curation applied?

Commented on VirtuProbe Studio · Jul 28, 2026

Hi DevHunt, I'm Jakub. I started VirtuProbe four years ago on evenings and weekends, after two decades building this kind of software professionally. I was sick of testing one signup flow across three separate tools: a request client for the API, a database client to check the row got written, and a mail viewer to confirm the email actually went out. Nothing let me say "take the userId this API returned and go check the record it created." So I built the thing that could. It's one workbench where a test chains across protocols. Each step's output feeds the next through extractors, so an HTTP response can drive an SMTP send, an LDAP lookup, a MySQL query, and so on. Eleven protocols today. When clicking through steps isn't enough, you can drop into Groovy and drive the whole thing programmatically. I wrote each protocol by hand against its RFC instead of wrapping a library, which is what makes the odd stuff work: raw wire-level requests, malformed-packet fuzzing, a proxy that can mock and rewrite traffic. The part I'd most like your opinion on is what shipped last week in 2.0. VirtuProbe now runs an MCP server, so your coding agent can connect and get 22 tools: inspect the workspace, create a probe, save and run a chain, run a suite, run a fuzz sweep, import a collection. It builds tests in your workspace instead of improvising its own. Three things I cared about getting right: - It inherits your tier. There is no separate agent licence and no second permission model, because the agent calls the same services the interface calls. - You approve which projects it may touch before it can write or run anything, and you can revoke it mid-run. Reading is free; writing and running need your consent. - It uses your stored credentials without ever receiving their values. An agent can point a saved credential at the environment you pinned for it, and nowhere else. And you can watch it work. Its runs stream into the app step by step, refused attempts show up with the reason they were refused, and a failed step is a link that opens that chain in the editor while the agent is still going. Whatever it builds is an ordinary chain in an ordinary bundle, not a second-class object. You don't start from a blank page either. There's a library of over a thousand ready-made tests and chains you can pull straight in, from everyday API checks to full security workflows. It stopped being a solo project a while ago. I've built a small team of excellent people around it, with real focus on UX and on testing, on the craft that takes a tool from working to polished. The full HTTP workbench is free forever, no account, and your data never leaves your machine. The other protocols, scripting and the proxy are the paid tiers. I'm here to answer your questions.