# Cairn

A general-purpose AI search engine for problems with a known starting point and goal but an unknown path between them, first validated on autonomous penetration testing.

Cairn is built around a directed search through a state space defined by a known origin and a defined goal, a shape shared by problems such as penetration testing, vulnerability research, mathematical proof, and CTF challenges. It defines no fixed roles or workflows: instead, an explicit fact-intent graph built on a blackboard architecture tracks confirmed Facts, declared but unexecuted Intents, and Hints of human judgment that agents absorb on their next read.

Worker agents run an observe-orient-decide-act loop over the shared graph rather than following predefined job descriptions, coordinating only through the shared board rather than direct communication. Three task types share one worker: an initial bootstrap attempt, a reasoning step that checks whether the goal is met and proposes next intents, and an exploration step that executes one intent and reports a finding. Penetration testing is the first domain the engine has been validated on.

Website: https://github.com/oritera/Cairn