"flows": a custom Markdown runtime for visualizing long-running agent loops
"Flows" is a custom Markdown runtime designed to visualize and manage long-running agent loops, providing a way to monitor and debug persistent AI agent processes through Markdown-based rendering.
Background
"Flows" is an open-source tool by developer Sam Leeney that renders Markdown documents as dynamic, visual interfaces for watching long-running AI agent loops in real time. Instead of static text, a "flow" shows each step of an agent's reasoning or execution as a live, updating diagram — useful for debugging, demos, or monitoring agents that take minutes or hours to complete. It sits in the growing ecosystem of "agent observability" tools, alongside projects like LangChain's LangSmith or Arize AI's Phoenix, but focuses on a lightweight, developer-friendly Markdown-first approach. The key pain point it addresses: as AI agents run complex multi-step tasks (e.g., web research, code generation, data processing), developers need to see what the agent is doing *while* it works, not just after it finishes.