Conversational process runtime.

Aflow is an independent research project exploring how AI agents run real processes: durable skills, long-horizon memory, event-driven work, and human judgment at the right boundaries. It exists to test new agentic patterns quickly, in a production-grade setting, and to learn from what breaks. If you're researching similar problems or curious about the architecture, reach out.

Access is limited to a few invited testers — bring-your-own-model-keys.