AI system that let's you supervise and direct research

So, I have been using almost every AI-research tool they are great but, after using them for a while I realized that sometimes the LLM adds things in the report that are just completely made up. Most of these tools are generating data that is not even true. I ended up building a tool myself to tackle this. I do not have the knowledge to solve the black box problem of AI but I definitely do know how to write orchestration that is very strict. So I developed this own orchestration with implementation of harness engineering, so instead of letting the agent roam around on the web freely it acts like a worker and the users become the supervisor so all the necessary decision making, research decisions, direction source verification is for the users. Instead of AI summaries the users get's mathematically filtered data obtained from the sources which makes the hallucination risk significantly. I got the version 1.0.1 validated by people already but I am wondering if my tool can be actually used in high authority workspaces where accuracy is a must . Would love brutal feedback from people who actually work with agentic systems.