LangKit is an open-source toolkit for monitoring language models. It extracts measurable signals from prompts and responses, such as readability and complexity, similarity between prompt and response, sentiment and toxicity. For security and privacy it counts regular-expression patterns to flag sensitive data such as card and phone numbers, compares text with known jailbreak and prompt-injection examples, and detects refusals. LangKit works with the whylogs data logging library and is installed as a Python package from PyPI. It is a library rather than a hosted service, so you run it in your own environment.
LangKit was developed by WhyLabs, a US company, and is licensed under Apache 2.0. Because the code runs in your own infrastructure, prompts and responses stay there unless you yourself send logs or statistics to an external service. The source code can be inspected and, if needed, forked. The repository does not say whether the project is actively maintained; WhyLabs has since also published the source code of its platform as open source. Check the maintenance status before relying on it in production.