@BiggumsMcHoney
Best video I’ve found on MCP. I think AI is such slop and most other channels just talk buzzwords with no real information. This helped me get a verbal overview before I review the docs at work. I learned with RAG that some people will just give theoretical explanations but have no understanding of implementation.
@shubha07m
Thank you for explaining things so clearly. Your channel is really appealing to me because of how well you break down complex LLM concepts!
@austinjohnathan4073
These videos have been a huge help to our business! Thank you!!
@pauliusztin
Great video, Shaw 🤟
@AIwithHonor
Thanks Shaw - you makes sense. Kind of like the MCP in tron
@ifycadeau
Perfect explanation! Thank you so much for the breakdown!!
@KaushikSharma-q2r
This video series is great. Would love to see more as you expand into newer topics ❤
@kassugebresellasie803
As always, great one Shaw!
@ChrisAuer
Great video. Clearly explained.
@CleanupStack
Great video. You earned a new subscriber!
@jawaidekram1288
Great Job!!!!
@jamaicauptown
Thanks this just taught a bunch of spammers to automate email deliveries :-|
@aeiou...
So basically, its a tool call that you made accessible for LLM GUI rather than integrating it into their API
@ppgab
So its basically tools but in a common format
@Qazi-it9mo
Shaw createss an assistant named 'AVA'. AVA: "Isn't it strange, to create something that hates you?"
@josephsylvan6511
Hey @ShawhinTalebi, As usual great content here :) I have been following your LLM and AI Agents playlists. I'd like you to talk about tool integration and retrieval, suppose I have thousands of tools...would it be like we end up writing all tool descriptions inside the mcp client? If not what's the approach the mcp client must follow? My understanding is that....these so call mcp clients are like some client connected to an llm....converts natural language and maybe do some semantic search based on tool description which might be stored in some vector db? And then we retrieve top k (~1-3 tools, maybe let user decide which tool) and try to execute those tools , here the natural language gets converted into specific tool template for execution....and then final output is passed back and maybe we get a summary of the execution process. So its more like RAG based MCP? And ofcourse each tool could itself be an AI model, maybe refine the repsonses...loop feedback and stuff. If this ideology is correct, then we're limited by the llm context and tokens? How do we design our own mcp clients?
@vitocorleone4139
When using whatsapp mcp, can we encrypt the data we send and send it, is it useful? I'm thinking of using whatsapp mcp for a long-term project. Or would it make more sense to use the meta api in terms of the user? Currently, the advantage of mcp is that it provides qr login to whatsapp.
@haneulkim4902
Great tutorial! One question, is there a way to run AVA in the background, like make it take care of my email without needing to prompting in claude
@ksh-v8x
5:21 can we call it drivers ?
@ShawhinTalebi