This product was not featured by Product Hunt yet. It will not be visible on their landing page and won't be ranked (cannot win product of the day regardless of upvotes).
Diffs two agent runs and shows which tool calls fired, how the arguments differed, what each run cost, and where the outputs diverged. Reads JSONL, OpenTelemetry, Langfuse and LangSmith exports. Run it in CI and fail the build on a dropped call.
I changed one line in a system prompt a few weeks ago and the agent quietly stopped making a refund call. It still wrote "your refund has been processed" in the final message, so the text looked fine, and the eval score moved from 0.82 to 0.79, which told me nothing. I found it by opening two trace files in split panes and scrolling until my eyes crossed.
That's the whole reason this exists. "The agent got worse" isn't a debuggable statement. Prompt edits, model swaps and framework bumps all shift behaviour without announcing it, and a score going down tells you that something moved, not what moved.
whatbroke takes two exported traces and diffs them. Which tool calls fired in each run, how the arguments differed, what each run cost and how long it took, and where the outputs actually diverged. It reads plain JSONL, plus OpenTelemetry, Langfuse and LangSmith exports. There's a --fail-on flag and a GitHub Action, so you can wire it into CI and break the build when a run loses a call it used to make.
It's a CLI, TypeScript, MIT, 139 tests. v0.5.0 went up yesterday.
Two things I actually want to know from you. What trace format are you on, because the importer only covers four right now. And is the diff output readable to someone who didn't write it, because I've been staring at it too long to tell.
npm i -g whatbroke-cli
No comment highlights available yet. Please check back later!
About whatbroke on Product Hunt
“Diff two agent runs and see exactly what changed”
whatbroke was submitted on Product Hunt and earned 4 upvotes and 1 comments, placing #93 on the daily leaderboard. Diffs two agent runs and shows which tool calls fired, how the arguments differed, what each run cost, and where the outputs diverged. Reads JSONL, OpenTelemetry, Langfuse and LangSmith exports. Run it in CI and fail the build on a dropped call.
whatbroke was featured in Open Source (68.7k followers), Developer Tools (517.5k followers), Artificial Intelligence (475.9k followers) and GitHub (41.4k followers) on Product Hunt. Together, these topics include over 234k products, making this a competitive space to launch in.
Who hunted whatbroke?
whatbroke was hunted by Arthi Arumugam. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how whatbroke stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Arthi Arumugam