Hacker News new | past | comments | ask | show | jobs | submit | jswny's comments login

Does anyone know how MCP servers would be used via the API?

I thought they ran locally only, so how would the OpenAI API connect to them when handing a request?


You'd use a client which runs locally to coordinate between the LLM/agent and the available tools, similarly to how it's described here https://modelcontextprotocol.io/quickstart/client

There are a variety of available clients documented here https://modelcontextprotocol.io/clients

If you haven't tried any of these yet, the first place to start is Claude Desktop. If you'd like to write your own agents, consider https://github.com/evalstate/fast-agent

EDIT: I may have misunderstood your question. If you're asking "how can I make an API call to OpenAI, and have OpenAI call an MCP server I'm running as part of generating its response to me", the answer is "you can't". You'll want a proxy API that you call which is actually an MCP client, responsible for coordinating between the MCP servers and the OpenAI API upstream agent.


You can run remote MCP servers and configure whatever client to use them. This should work even via OpenAI's API(perhaps not yet, but it's just another 'tool' to call)

https://blog.cloudflare.com/remote-model-context-protocol-se...


command mode is for local, and SSE for remote


Zed already has built in task running… you can use the same thing to call anything that you want like a command to run your project. You can even add custom keybindings to them

https://zed.dev/docs/tasks


I’m trying to find where I can have it as a button so every time I want to run / debug the code, I can press on it instead lf having it as a shortcut but I haven’t found a way.


I agree. I actually use the AI features in Zed a lot, but there are things I really wish they would prioritize.

For example this issue that’s been open for about a year: https://github.com/zed-industries/zed/issues/10122

Editing large files is an incredibly common use case for an editor


The pricing is completely prohibitive. It’s a shame, looks cool but I’m not gonna bother with a pricing model that’s clearly telling me to go look elsewhere as a solo dev


I would try this but I want to be able to use it on mobile too which requires hosting. I don’t see how valuable an LLM interface can be if I can only use it on desktop


How can this company possibly support maintaining 2 browsers… I don’t hate the idea they are going for but I swapped off of Arc the moment I heard they were trying to do yet another browser


The thing about local first syncing options like this is that they mostly do not work on mobile. For example iPhones cannot handle dropbox syncing random text files in the background as a regular filesystem for an app to deal with.

Not saying that's not iPhone's fault, but I doubt any of this works on that platform


I've been a happy user of a PWA doing local sync. That said, the data it needs to sync can fit in localStorage.

Not affiliated in anyway, but the app is http://projectionlab.com/ and it allows you to choose between json import/export, localStorage sync, and server-based sync as desired. Since it has an easy to use import/export, sync with some other cloud provider on iOS is basically just a matter of "saving the file," since iOS lets you do background sync of authorized Files providers.

Even though it's a web app, being able to download the page and then after run it entirely offline in a fresh browser window each time, built a lot of trust with me, to the point where I mostly run it with localStorage enabled and only occasionally check its online/offline behavior anymore.


not working on mobile (phones) is fine in my book as long as it does support laptops


Just hit “not interested” on those, the YouTube algorithm is really good at not showing you stuff you tell it not to


The issue is that there’s an infinite number of channels that do Minecraft videos.

It’s like playing whack-a-trying to hide all the Minecraft channels.

I want to hide the topic, not an individual channel.


There are two options. "Not interested" nudges the algorithm away from the topic while "Don't recommend <type> from this channel" immediately hides the channel from recommendations.

If your nudges don't feel effective enough, you can also manage your view history. Removing any Minecraft videos you may have accidentally watched from your history will help hide them from future recommendations.


I’ve tried all of the above and Minecraft keeps coming up.

I think the engagement numbers are tracked per user, and that’s what they recommend off of. Data outside the scope of the listed user history.

It’s insanity. And infuriating.

I want a diverse set of recommendations, not another 1,000 Minecraft videos.


File feedback in the app. My understanding is that people in charge of the recommendation system actually look at it and try to use it to make the system better.


but I want it to hide on the client side.


Out of curiosity, why?


I’m not OP but I use YouTube a lot. I would assume two reasons:

Some people would rather scalp themselves than create a google account (valid)

Even when logged in and using the “Not interested” or the “don’t recommend channel” button, YouTube will sometimes straight up ignore you.


Is Turbo supposed to be a better version of GPT4? I thought it was just supposed to be a faster, slightly less accurate version


Static typing for Elixir is in active development: https://elixir-lang.org/blog/2023/09/20/strong-arrows-gradua...

Yes, it is currently dynamic and yes, that means you won't get much int he way of type information from the language itself, it's not currently a statically typed language.


dynamic but still strongly typed


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: