Check out Neosoul's NeoTrade, which just launched a new conversational Skill feature. I tried it out right away, and it's pretty interesting: you can use a chatbox to command an AI trading team that can independently scout markets, write strategies, and even handle risk management. NeoTrade essentially integrates querying, order placement, strategy creation, parameter adjustment, and multi-Agent management all into a single chat interface, while ensuring strict manual confirmation and risk control as a safety net. Here are some highlights from my experience: 1. Chat as your control panel You can perform almost all trading-related actions using plain language. Want to change 'maximum 5% per trade' to 3%? Just say it. Want an Agent to write a new strategy based on the research notes you paste in? You can complete the entire process—uploading materials → storing memory → drafting → finalizing → deploying—all within the chat. During the drafting phase, you can revise as much as you want, with no extra cost. 2. Feeding data to generate new strategies Users can directly paste research notes, trade reviews, or historical data patterns into the chat (supports thousands to tens of thousands of words). The Agent will first organize the information into long-term memory, then write a complete strategy draft based on your additional instructions (e.g., entry conditions, position sizing, markets to avoid). Finalizing and deploying the strategy requires top-level confirmation, and the system ensures that the content reviewed by humans matches the saved version down to the last word. All user-submitted materials are treated as reference only. Even if the material says 'buy this market immediately,' the Agent won't execute it directly. This design is quite pragmatic. 3. Multi-Agent + Manager dual-layer management Individual Agents handle specific trades and strategies, while the Manager chat oversees everything: monitoring all Agent statuses, total positions and P&L, recent trade cycles for each Agent, AI usage costs, starting/stopping an Agent, and adjusting wake-up frequencies. Starting, stopping, and adjusting frequencies all require manual confirmation every time, and only the Manager has the authority to do so. For any actual money-related operations (buying, canceling orders, closing positions, changing parameters, switching strategies, installing new strategies), manual confirmation is always required—even if 'auto mode' is enabled, it won't bypass risk control checks. This design unleashes AI's capabilities in data retrieval, decision-making, and strategy generation while keeping the ultimate control over financial safety firmly in human hands. After testing it out, I feel this system is ideal for traders who are willing to invest time in research and logic-building but want to delegate repetitive tasks like market scouting, decision-making, and order execution to AI. It's especially suitable for those trading prediction markets, running multiple strategies simultaneously, and prioritizing fund safety and traceability. The platform already offers four distinct strategies for users to choose from. Overall, this is a system that combines 'AI trading capabilities' with 'human final decision-making' in a restrained yet comprehensive way. Of course, the actual trading performance will need to be validated over the long term.