I feel like I slept for a day and now MCPs are everywhere... I don't know what MCPs are and at this point I'm too afraid to ask.
and then the LLM model will ask the MCP server to call the functions, check the result, call the next function if needed, etc
Right now if you go to ChatGPT you can't really tell it "open Google maps with my account, search for bike shops near NYC, and grab their phone numbers", because all he can do is reply in text or make images
with a "browser MCP" it is now possible: ChatGPT has a way to tell your browser "open Google maps", "show me a screenshot", "click at that position", etc
It seems strange to me to focus on this sort of standard well in advance of models being reliable enough to, ya know, actually be able perform these operations on behalf of the user with any sort of strong reliability that you would need for widespread adoption to be successful.
Cryptocurrency "if you build it they'll come" vibes.
EDIT: Don't get me wrong, the benchmark scores are indeed higher, but in my personal experience, LLMs make as many mistakes as they did before, still too unreliable to use for cases where you actually need a factually correct answer.
Believe me. It's not there yet.
I was referring more broadly to ClaudePlaysPokemon, a twitch stream where claude is given tool calling into a Gameboy Color emulator in order to try to play Pokemon. It has slowly made progress and i recommend looking at the stream to see just how flawed LLM's are currently for even the shortest of timelines w.r.t. planning.
I compared the two because the tool calling API here is a similar enough to an MCP configuration with the same hooks/tools (happy to be corrected on that though)
Is this what 'calling' is?
So im not sure id give up the sum total progress of the automobile just because the first decade was a bad one
The article also reeks of LLM ironically
https://invariantlabs.ai/blog/mcp-security-notification-tool...
Yes, MCP is a way to streamline giving LLMs ability to run arbitrary code on your machine, however indirectly. It's meant to be used on "your side of the airlock", where you trust the things that run. Obviously it's too powerful for it to be used with third-party tools you neither trust nor control; it's not that different than downloading random binaries from the Internet.
I suppose it's good to spell out the risks, but it doesn't make sense blaming MCP itself, because those risks are fundamental aspects of the features it provides.
It introduces a substantial set of novel failure modes, like cross-tool shadowing, which aren't obvious to most folks. Making use of any externally developed tooling — even open source tools on internal architecture — requires more careful consideration and analysis than most would expect. Despite the warnings, there will certainly be major breaches on these lines.