super great tutorial , excuse my many question at each blog post
but i wonder , if i use the mcp server with steamtable http and host it on my company server , does the client implementation aka agent will be different ? also what if i used stdio on my server i think the client must be on the same server , right ?
Yes so the MCP protocol currently has 2 transport mechanisms.
stdio - the server starts as a subprocess of the client, so same memory space, same cpu space, works only on the same machine.
streamable-http - is destined for remote setups. Client and server will talk remotely.
The Agent implementation won’t be different, just make sure that the host running your Agent(mcp client) can access the IP:PORT of the server hosting the MCP Server.
If from the machine you run your agent, you can do ‘ping <ip:port of mcp server> and you get responses back - you’re good to go.
Otherwise, you ll have to set the connectivity, public ips, or whitelist ports between machines.
Hi there! I’m a tech-adjacent human who recently started a Python class (zero coding experience) mostly because I want to peek behind the prompting curtain. Would you say Python is the best first step toward understanding AI and machine learning, especially if I’m hoping to eventually build my own app?
Hey, totally! - Starting with Python is the best way to get into ML, as it's the dominant language in the AI field.
I wouldn't recommend any other path than starting to learn Python.
The majority of high-level libraries/frameworks for AI are built in Python, with only a small portion of them being built in Go, Rust, C++, or other languages.
Dude, this article is AMAZING! 😍
super great tutorial , excuse my many question at each blog post
but i wonder , if i use the mcp server with steamtable http and host it on my company server , does the client implementation aka agent will be different ? also what if i used stdio on my server i think the client must be on the same server , right ?
could you wrap it up please
big thanks <3333
Hey,
Yes so the MCP protocol currently has 2 transport mechanisms.
stdio - the server starts as a subprocess of the client, so same memory space, same cpu space, works only on the same machine.
streamable-http - is destined for remote setups. Client and server will talk remotely.
The Agent implementation won’t be different, just make sure that the host running your Agent(mcp client) can access the IP:PORT of the server hosting the MCP Server.
If from the machine you run your agent, you can do ‘ping <ip:port of mcp server> and you get responses back - you’re good to go.
Otherwise, you ll have to set the connectivity, public ips, or whitelist ports between machines.
Hi there! I’m a tech-adjacent human who recently started a Python class (zero coding experience) mostly because I want to peek behind the prompting curtain. Would you say Python is the best first step toward understanding AI and machine learning, especially if I’m hoping to eventually build my own app?
Hey, totally! - Starting with Python is the best way to get into ML, as it's the dominant language in the AI field.
I wouldn't recommend any other path than starting to learn Python.
The majority of high-level libraries/frameworks for AI are built in Python, with only a small portion of them being built in Go, Rust, C++, or other languages.