Hope this helps someone to understand:
1. How MCP works as standalone client and server
2. How MCP works with LLM (Claude to be exact, others should be the same)
I made each commit atomic so you can review the code to see how it works. The chat cli demo commit actually mimics interacting with LLMs so you can play with it without having Anthropic API key.
Thanks for the post. I can related to what you posted and would like to offer few perspectives that might help you.
1. Start with accepting the fact that you are the lead developer who is responsible for the long term success of the project thus the team. This is really important in that it puts the project and team as the same level.
2. Start mentoring and delegating more core pieces to others in the team. Allow them to make mistakes of their own so they can see the bad and appreciate the good. As long as their work is good enough, let them take more ownership so they can grow be the lead developer if you are ever to move on.
3. The last point is to always remind yourself of the fact that the project will eventually continue regardless of you being there or not. As a lead, it is your responsibility that it will continue smoothly after you move on to other things. As long as you keep doing "one more thing you got to fix", you will accumulate more responsibilities that will eventually burn you out. So for the good of the project, the team and yourself, you should start working towards a scenario where either the project or the team does not need to depend on you to be able to move forward, which will eventually happen whether you like it or not.
Once you get proxy url as shown in the bash script, you can post to it whatever info they require (most of the time, it is just an email or zip code etc). The web gui pop up is just browser window opened by your OS. If you happen to use BSD/Linux, you are most likely will not see the popup. It is surely the case with OpenBSD.