When the concept of agents first emerged, many popular open-source implementations and startups appeared. Eventually, Claude Code became the standard, bringing about an ecosystem that includes MCP, Skill, and others. Proactive agents, or called agentic loops, also started to explode in the open-source community, but it probably took about a year for large model companies to develop truly usable products and ecosystems. I estimate that on the enterprise side, it's Anthropic, while for individuals, it depends on whether OpenAI's product approach is reasonable. Google itself might still be behind, after all, a truly excellent agent ecosystem has not yet been established.