What is nitpick?

Nitpick is an open source tool that automates the early, repetitive parts of handling production errors. It watches error logs coming from Datadog, groups together errors that are really the same underlying problem happening again and again, and then hands each new distinct error to Claude, Anthropic's AI coding assistant, to investigate. For each new error, Claude reads the relevant parts of your codebase, works out a likely root cause, and Nitpick turns that into a written root cause analysis document plus a filed issue in GitHub or Linear. If a fix looks safe and small enough, typically under 250 lines of code, Nitpick can also create an isolated copy of the repository, apply the fix there, and open it as a pull request for a human to review. It never pushes a fix directly to a live checkout, and automatic fixes are switched off by default until someone deliberately turns that setting on for a given part of the codebase. According to the README, a real deployment has been running this continuously for about two months at a voice AI company, processing over 40,000 raw log entries down to a few hundred distinct recurring problems, filing over a thousand issues automatically, and opening 21 fix pull requests, six of which were merged. Nitpick runs locally rather than as a hosted service, so error data and source code stay under your control, and only the specific records needed to file an issue or write a document are sent to outside providers. Right now it supports Datadog as its error source, with GitHub or Linear for issue tracking and Markdown or Notion for documentation, and the project describes more sources and integrations as planned future work. This is aimed at engineering teams who want repeated production errors turned into reviewable, actionable work instead of getting lost in noisy alert channels, without giving an AI system the ability to change production code unsupervised.