Reacknadam/openagent ? reverse-engineered prompt

Build me a self hosted AI coding agent web app that can take a task, break it into steps, use tools, and keep working until it checks the result. I want a clean chat style interface with real time updates, a history panel, a terminal style output area, and a place to manage models and tools. It should support both local and cloud models, let me switch between providers easily, and work with common options like OpenAI, Gemini, Claude, Ollama, Groq, DeepSeek, OpenRouter, LM Studio, and custom compatible APIs.

The agent should plan, reason, call tools, reflect on the results, and validate the final answer instead of just replying once. Add memory for conversations and ongoing work so longer tasks do not get lost. Make it production ready, easy to configure with environment variables, and safe enough to use on my own machine. If you need to look up current docs online while building it, go ahead.