https://github.com/runagent-dev/runagent/tree/main/examples/Vibe_Agent_Builder_gpt_5

I created a tool that lets you write a prompt like “build a research report agent in LlamaIndex” or in any framework and automatically generates and deploys the agent locally, complete with a testing page. Powered by OpenAI’s GPT-5-mini, it’s stable (98% success rate in my tests) and works with frameworks like Agno and LlamaIndex. Previous GPT models struggled with this kind of tasks, but this handles them well.