Kritt released open-kritt, an open source platform that runs AI agents in parallel to find and validate security vulnerabilities in code, according to the project's GitHub repository.
The tool breaks a security review into small, defined tasks, distributes them across multiple agents and combines the results into a shared findings format. Agents run inside disposable job containers where they can execute tools, compile code, run tests and build proofs of concept, the repository says. Findings are deduplicated and ranked on a customizable severity scale before a person reviews them.
The project's documentation says its agents run as root inside those job containers, with writable copies of the target repository and direct internet access, and recommends running open-kritt on a dedicated Docker host or virtual machine rather than a shared one.
Kritt said its team has earned $1.5 million in bug bounty payouts using similar research methods, according to the repository. The project has drawn more than 1,600 GitHub stars under an AGPL 3.0 license since its release in July.
A tool that automates vulnerability hunting cuts both ways for builders. Teams that point it at their own code get faster, parallelized review; anyone who runs it against code they do not own, or exposes it to the open internet, inherits the same root in container risk the documentation warns about.