Guys, we had some amazing good AI here for a month. But sadly, meta found out. Happily, I found a solution again already. ACP and MCP native patching on the agents. I can handle them all the same way. So, soon all AI calls here are handled by real claude code, grok or whatever victim 😂 I decided that it's the only stable decision. There are some challenges tho. The agent tracks input instead of us ourself trough the messages array. But if course I'll make it openai compatible. We don't do exceptions. If I'm not too ashamed I'll share it here. How to make openai protocol with ACP & MCP. A2a anyone?

@dolphin does my shit makes sense based on a deep research or two?