Apparently you aren’t supposed to create a while loop without an exit either, but
@yoheinakajima
-
MCPs Tool Management and Automated Filtering for AI Agents
By
–
Yes! I want to have a massive list of tools that are embedded and filtered down to likely ones automatically but I don’t think that’s easy with MCPs…. but maybe you can?
-
Codex AI Tool for Legacy Project Development and Agent Orchestration
By
–
unlike other AI coding tools, I’m excited to try throwing codex at some older projects seems like this is just the “execution agent”, and once there’s cohesive orchestration on top of this…
-
Managing Ads from IDE Development Environment
By
–
i think this is more managing ads from an IDE, but well… you never know
-
AI-Powered Marketing Automation Launches From IDE with Product Context
By
–
love this new DX. launch automated marketing/ads directly from your IDE w full context of your product. https://t.co/lYGmNnPmUA
— Yohei (@yoheinakajima) 19 mai 2025love this new DX. launch automated marketing/ads directly from your IDE w full context of your product.
-
MCP Architecture Best Practices for Personal Projects
By
–
This seems like the "right" way to do it. I guess for small personal projects, a single MCP might work but it's basically the lazy way.
-
Bundling Tools by Use Case: Multi-Purpose Tool Considerations
By
–
I see. Seems like bundling by use case might be nice. But then some tools might be useful in multiple use cases…
-
MCP Strategy and Its Distance from Autonomous Agents
By
–
I have portfolio companies asking me about MCP strategy so trying to get a feel for it, but I agree that it feels further from autonomous agents 🙂
-
Tool-Specific MCP Open Source Repositories Discussion
By
–
Makes sense. Saw a lot of tool specific MCP open source repos which I think confused me (though it makes sense why the open source repos are tool specific).
-
MCP Server Architecture: Single Large vs Multiple Specialized
By
–
this might seem like a silly question but if you're setting up remote MCP server(s) for your own AI projects, do you really need multiple or can you just create one large MCP server with many tools or does that defeat the purpose somehow?
— Yohei (@yoheinakajima) 19 mai 2025
(relevant experiment in video) pic.twitter.com/B9h6ntOQUXthis might seem like a silly question but if you're setting up remote MCP server(s) for your own AI projects, do you really need multiple or can you just create one large MCP server with many tools or does that defeat the purpose somehow? (relevant experiment in video)