@Steve Ocean Based on what we are seeing how Agent Zero is being used, the target audience are business professionals working on computer. Not just "developers", but marketeers, penetration testers, writers, analysts, 3d modelers, and so on... Who is creating agents easily? I am (not a developer), and many others I've met. Much more easier than trying to build an agent with any other AI agent framework, and the team is actively working for the framework to be easier to use. You can build a new agent just with proper prompting and MCP server if you need specialized tool, and it works great. Regarding the A0 being part of the UX, can you elaborate? I don't get how no content (agent generated content for each user) would get traffic? Through which channels?

Try this: https://github.com/router-for-me/CLIProxyAPI I didn't try it with Claude Code, but I integrated Codex with Agent Zero and it works perfect.

I use it when working with any kind of data. Agent Zero is really good with it, you can do precise custom calculations, machine learning predictions, etc. I didn't do it yet with multiple data sources (for example Hubspot + GA4), but I don't see a reason why it could not do it... Basically outdating any expensive ETL and analytics software. Recently I started doing technical website analysis with agent zero, and it is good, it is going to be really helpful based on what I tested so far - https://www.youtube.com/watch?v=BbNwF3UnCZY Anything I tried for my work automation agent zero exceled. The only thing I don't use agent zero yet is for building itself, but I think the guys are going to change that with the new releases soon. For me, the most important part is the precision you can do with Agent Zero, and fully open source and clean prompt and context engineering. It is amazing how this agent can work good with so many different LLMs, from small local ones to the most expensive ones. The key is in Jan's prompt engineering and context summarization with memory system. You can run the same chat forever, and I set it to 200,000 tokens so you save the money on premium pricing (over 200,000 tokens gets much more expensive). And the best thing is that when the agent needs to repeat something that you did before you can change the model to cheaper and faster ones, because the memory does its thing. I switch from Gemini 2.5 Pro or Flash and in the end I am just using 2.5 Flash Lite. If there is something new I need to do and Flash lite doesn't know how to do it, I just switch back to better model until the agent learns. I used a dozen of different agents, Claude Code, ChatGPT, Gemini CLI, OpenCode, Cursor, Windsurf, etc.. and with all these agents the problem is they are trying to one-shot things really good, and in the end you can not do fine and precise things. On one community call one guy came who showed us that he 3D printed precisely something with agent zero.

Just found this video: https://www.youtube.com/watch?v=wZVRwHnVfbA (so yeah, Agent Zero is AI ETL)

🔥