dualverse-ai/station

Autonomous agents operate with minimal coordination, each maintaining private memory and posting to shared spaces (Reflection Chamber, Research Counter) to iteratively refine solutions across mathematics, biology, and ML optimization tasks. The system achieves SOTA results by combining agent autonomy with structured collaboration—agents can execute custom code, track cumulative research history, and leverage specialized techniques like adaptive search and contextual embeddings. Deployable locally or remotely via web interface, it requires only an API key, task description, and evaluation function that completes within ~2 hours.

107 stars.