Revalvo
Product Hunt Lokesh
Revalvo launched as a local-first workbench for prompt engineering and LLM evals. It tests prompts across models in parallel, and keeps your API keys in the browser.
Based on reporting by Product Hunt, Lokesh — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Revalvo is trying to make prompt work feel more like software work. The new tool is a local-first workbench for prompt engineering and LLM evaluation, built to let people run the same prompt against every model in parallel and compare what comes back.
The pitch is practical rather than flashy. Revalvo comes with 40 built-in evaluators, so responses can be scored without stitching together a separate stack. Prompts can also be versioned like code, which is the sort of feature that sounds minor until a team has to track down which wording caused a result to drift.
It also supports batch testing on datasets before anything reaches production. That matters because prompt tinkering gets messy fast once a project moves beyond one-off experiments and into repeatable checks.
And then there’s the local-first part, which is probably the sharpest line in the whole product. No account. No hosted database. API keys stay in the browser. That’s a simple promise, and for teams nervous about where their prompt data ends up, it may be the one that gets attention first.
My take — AI-written commentary, not fact-checked reporting
Revalvo is sensible in a field that often mistakes more UI for more rigor. Keeping the whole thing local and browser-based is the right instinct; prompt tooling shouldn’t need a trust fall. The real tell is whether teams use the evaluators before production, or just admire them and keep shipping vibes.
Read more about this at: Product Hunt