Cairn is a general-purpose state-space search engine that uses AI agents to navigate from a known origin to a defined goal through an unknown problem space.
Cairn addresses the problem of directed exploration in domains where the starting point and success condition are clear but the path between them is unknown. It uses a blackboard architecture built on three primitives: Facts (confirmed findings written to a shared board), Intents (declared directions of exploration), and Hints (human guidance injected at runtime). Agent workers operate in an OODA loop—observing the full fact-intent graph, orienting to the current state, deciding on next intents, and acting to explore—then writing their findings back as new Facts. Workers have no fixed roles and coordinate exclusively through the shared board via stigmergy, with no direct communication between agents. Tasks are generated dynamically from the graph's current state rather than from predefined job descriptions.
The tool is validated first on autonomous penetration testing, where the structure of the problem—known target, defined objective like obtaining shell access, unknown exploitation path—maps directly to the engine's design. However, the architecture is intentionally general and applies to other domains with similar structure, such as vulnerability research, mathematical proof, and CTF challenges. Cairn defines no roles and no workflows; it requires only an origin, a goal, and a way to recognize success.
The project shows active development with ongoing refinement of the core search and reasoning mechanisms. The codebase demonstrates iterative work on agent coordination and fact-intent graph evolution. Community engagement is present through Discord and social channels, indicating sustained interest in the project's direction.
