161–170 of 348 posts
Earlier quoted context omitted.
You need something like OAuth because you don't want your end users generating API keys for every service they want to use via LLM.
Maybe we should though
Just to add one piece of clarification - the comment around authorization is a bit out-of-date. We've worked closely with Anthropic and the broader security community to update that part of MCP and implement a proper separation between resource server (RS) and authorization server (AS) when it comes to roles. You can see this spec in draft[1] (it will be there until a new protocol version is ratified). [1]: https://m…
It's setting off all kinds of alarm bells for me, and I'm wondering if I'm on to something or if my LLM-detector alarms are miscalibrated.
Earlier quoted context omitted.
You'll come around to my perspective in time. Don't take it personally. This generation isn't any worse than prior ones. We go through this shit every time the tech industry turns over.
The people who built MCP are seasoned software engineers, as are most folks who work for these labs. What are you even on about?
Earlier quoted context omitted.
The AI houses buying up the entire market of GPUs. Have you heard about them?
This is paranoid drivel.... Tell me which is more likely. 1. There is a cabal of companies painstakingly working together to make the most convoluted software possible from scratch so they can dominate the market. or 2. A few people threw together a bit of code to attempt to get something working without any deep engineering or systematic view of what they were trying to accomplish, getting something to work well eno…
This is almost certainly because they're all using LLMs to write the documentation, which is still a very bad idea. The MCP spec [0] has LLM fingerprints all over it.
In fact, misusing LLMs to build a spec is much worse than misusing them to avoid writing good docs because when it comes to specifications and RFCs the process of writing the spec is half the point. You're not just trying to get a reasonable output document at the end (which they didn't get anyway—just try reading it!), you're trying to figure out all the ways your current thinking is flawed, inadequate, and incomplete. You're reading it critically and identifying edge cases and massaging the spec until it answers every question that the humans designing the spec and the community surrounding it have.
Which means in the end the biggest tell that the MCP spec is the product of LLMs isn't that it's somewhat incoherent or that it's composed entirely of bullet lists or that it has that uniquely bland style: it's that it shows every sign of having had very little human thought put into it relative to what we'd expect from a major specification.
[0] https://modelcontextprotocol.io/specification/2025-03-26
Earlier quoted context omitted.
Speak for yourself. I see the majority of work being done by professional software engineers.
Any popular examples to support your claim? My claim is supported by the post article and many points there, for example. Another example is my own experience working with python ecosystem and ai/ml libraries in particular. With rare exceptions (like pandas) it is mostly garbage from DevX perspective (in comparison of course). But I admit my exposure is very limited. I don’t work in ai area professionally (which is a…
Earlier quoted context omitted.
Yes, the protocol seems fine to me in and of itself. It's the transport portion that seems to be a dumpster fire on the HTTP side of things.
This article feels like an old timer who knows WebSockets just doesn't want to learn what SSE is. I support the decision to ditch WebSockets because WebSockets would only add extra bloat and complexity to your server, whereas SSE is just HTTP. I don't understand though why have "stdio" transport if you could just run an HTTP server locally.
That all said, the real underlying mistake here isn't the choice of SSE... it is trying to use JSON-RPC -- a protocol which very explicitly and very proudly is supposed to be stateless -- and to then use it in a way that is stateful in a ton of ways that aren't ignorable, which in turn causes all of this other insanity. If they had correctly factored out the state and not incorrectly attempted to pretend JSON-RPC was capable of state (which might have been more obvious if they used an off-the-shelf JSON-RPC library in their initial implementation, which clearly isn't ever going to be possible with what they threw together), they wouldn't have caused any of this mess, and the question about the transport wouldn't even be an issue.
Earlier quoted context omitted.
To this date I have not found a good explanation what an MCP is. What is it in old dev language?
old dev language is deterministic, llm in the loop now so the language is stochastic.
Earlier quoted context omitted.
Any popular examples to support your claim? My claim is supported by the post article and many points there, for example. Another example is my own experience working with python ecosystem and ai/ml libraries in particular. With rare exceptions (like pandas) it is mostly garbage from DevX perspective (in comparison of course). But I admit my exposure is very limited. I don’t work in ai area professionally (which is a…
pytorch, tensorflow, numpy there are quite a few examples ai/ml has been steadily more commodetized, so it's far from only being developed by mathematicians. Hence every highschools student and his mother has an AI startup now. (And I'm not even mad, it's actually very exciting to see what people come up with nowadays)