I'm a photographer. I built a DSL for multi-agent workflows (github.com)

A photographer has introduced a declarative domain-specific language (DSL) called AgentFlow, designed to create multi-agent AI workflows without requiring coding knowledge. Users can write workflows in clean, readable `.aflow` files, which are compiled into MCP (Multi-Agent Collaboration Protocol) tools, streamlining integration with AI models like Claude, OpenRouter, and Ollama. The language supports iterative processes where different agents—each powered by distinct AI models—collaborate seamlessly through structured output. For instance, in a blog post workflow, agents are assigned specific tasks including research, writing, and editing, with built-in revision loops to enhance quality. This innovation is significant for the AI/ML community as it democratizes access to building complex workflows, making them more accessible for non-programmers while reducing boilerplate code. Compared to traditional approaches, AgentFlow allows for easier configuration and model assignment, with workflows that are Git-friendly and reviewable without extensive technical knowledge. The automatic MCP integration enables fast deployment of custom workflows as tools within Claude Code, showcasing a significant step in enhancing collaboration among AI systems and optimizing workflow management.