for teams that give their AI coding assistant rules
Harnessbench remembers the real tasks you give Claude Code or Codex. Whenever you change the assistant's instructions, it runs those same tasks again and tells you, in plain terms, what got better and what got worse.
works with
how it works
Every task you give Claude Code is saved on its own. You don't change anything about how you work.
Edit CLAUDE.md, add a hook, add a skill. Every saved task runs again: once with the old rules, once with the new.
A yes or a no, in plain words, with the exact line that caused it. Not a dashboard to interpret.
one real example
The rule was "always ask before editing a file the task didn't mention." The task was renaming a module that touches eleven files.
The rule did exactly what it said. It also turned a ten-minute job into one that never finishes without you. Harnessbench shows you that before the rule goes in.
more examples
how it scores
Your own tests run first, so the score starts from facts. Then a reviewer model reads what the assistant did and answers these, with its reasoning written down.
pricing
The CLI, the replays, the grading and the local dashboard are free and open source. Plans add what teams need: history, shared results, and a check on every pull request.
or $39 per repo / month
questions
No. Everything runs on your machine, or on a build server you already control, using the assistant subscription you already have. Harnessbench never sees your code.
Harnessbench answers for you, the same way you did when the task was first saved. If you never got asked the first time, it says "go ahead". Every question it had to answer shows up in the report, because an assistant that suddenly asks a lot more questions is usually the whole story.
The score starts from things that aren't opinions: did your tests pass, did it touch files it shouldn't. Only after that does a reviewer model read along and answer the four questions, and it has to write down why. You can read its reasoning and overrule it.
About as long as your slowest saved task, because they run at the same time. The twelve-task example above took eleven minutes.
get started