EvoAgentX is an open-source framework for building, evaluating, and evolving LLM-based agents and agentic workflows in an automated, modular, and goal-driven manner. The project introduces a self-evolving agent ecosystem where AI agents can be constructed, assessed, and optimized through iterative feedback loops, similar to continuous integration in software development. From a single prompt, EvoAgentX can build structured multi-agent workflows tailored to specific tasks. The framework then improves these workflows using self-evolving algorithms that optimize agent behavior over time. Memory modules provide both short-term and long-term memory capabilities, enabling agents to remember, reflect, and improve across interactions. Human-in-the-Loop support allows users to step in at checkpoints to review or guide workflows and step out again. EvoAgentX integrates with multiple LLM providers including OpenAI, Anthropic Claude, DeepSeek, Qwen, and others through LiteLLM, SiliconFlow, and OpenRouter. The framework supports RAG-enhanced agents, tool-augmented workflows, and natural language processing pipelines. Presented as a demo paper at EMNLP 2025, EvoAgentX represents a paradigm shift from static prompt chaining to dynamic, goal-driven agent workflows that continuously self-improve.