[HN Gopher] MCP in LM Studio
       ___________________________________________________________________
        
       MCP in LM Studio
        
       Author : yags
       Score  : 228 points
       Date   : 2025-06-25 17:27 UTC (1 days ago)
        
 (HTM) web link (lmstudio.ai)
 (TXT) w3m dump (lmstudio.ai)
        
       | chisleu wrote:
       | Just ordered a $12k mac studio w/ 512GB of integrated RAM.
       | 
       | Can't wait for it to arrive and crank up LM Studio. It's
       | literally the first install. I'm going to download it with
       | safari.
       | 
       | LM Studio is newish, and it's not a perfect interface yet, but
       | it's fantastic at what it does which is bring local LLMs to the
       | masses w/o them having to know much.
       | 
       | There is another project that people should be aware of:
       | https://github.com/exo-explore/exo
       | 
       | Exo is this radically cool tool that automatically clusters all
       | hosts on your network running Exo and uses their combined GPUs
       | for increased throughput.
       | 
       | Like HPC environments, you are going to need ultra fast
       | interconnects, but it's just IP based.
        
         | dchest wrote:
         | I'm using it on MacBook Air M1 / 8 GB RAM with Qwen3-4B to
         | generate summaries and tags for my vibe-coded Bloomberg
         | Terminal-style RSS reader :-) It works fine (the laptop gets
         | hot and slow, but fine).
         | 
         | Probably should just use llama.cpp server/ollama and not waste
         | a gig of memory on Electron, but I like GUIs.
        
           | minimaxir wrote:
           | 8 GB of RAM with local LLMs in general is iffy: a 8-bit
           | quantized Qwen3-4B is 4.2GB on disk and likely more in
           | memory. 16 GB is usually the minimum to be able to run decent
           | models without compromising on heavy quantization.
        
             | hnuser123456 wrote:
             | But 8GB of Apple RAM is 16GB of normal RAM.
             | 
             | https://www.pcgamer.com/apple-vp-says-8gb-ram-on-a-
             | macbook-p...
        
               | arrty88 wrote:
               | I concur. I just upgraded from m1 air with 8gb to m4 with
               | 24gb. Excited to run bigger models.
        
               | diggan wrote:
               | > m4 with 24gb
               | 
               | Wow, that is probably analogous to 48GB on other systems
               | then, if we were to ask an Apple VP?
        
               | minimaxir wrote:
               | Interestingly it was AI (Apple Intelligence) that was the
               | primary reason Apple abandoned that hedge.
        
             | dchest wrote:
             | It's 4-bit quantized (Q4_K_M, 2.5 GB) and still works well
             | for this task. It's amazing. I've been running various
             | small models on this 8 GB Air since the first Llama and
             | GPT-J, and they improved so much!
             | 
             | macOS virtual memory works well on swapping in and out
             | stuff to SSD.
        
         | karmakaze wrote:
         | Nice. Ironically well suited for non-Apple Intelligence.
        
         | incognito124 wrote:
         | > I'm going to download it with Safari
         | 
         | Oof you were NOT joking
        
           | noman-land wrote:
           | Safari to download LM Studio. LM Studio to download models.
           | Models to download Firefox.
        
             | teaearlgraycold wrote:
             | The modern ninite
        
         | sneak wrote:
         | I already got one of these. I'm spoiled by Claude 4 Opus; local
         | LLMs are slower and lower quality.
         | 
         | I haven't been using it much. All it has on it is LM Studio,
         | Ollama, and Stats.app.
         | 
         | > _Can 't wait for it to arrive and crank up LM Studio. It's
         | literally the first install. I'm going to download it with
         | safari._
         | 
         | lol, yup. same.
        
           | chisleu wrote:
           | Yup, I'm spoiled by Claude 3.7 Sonnet right now. I had to
           | stop using opus for plan mode in my Agent because it is just
           | so expensive. I'm using Gemini 2.5 pro for that now.
           | 
           | I'm considering ordering one of these today: https://www.newe
           | gg.com/p/N82E16816139451?Item=N82E1681613945...
           | 
           | It looks like it will hold 5 GPUs with a single slot open for
           | infiniband
           | 
           | Then local models might be lower quality, but it won't be
           | slow! :)
        
             | kristopolous wrote:
             | The GPUs are the hard things to find unless you want to pay
             | like 50% markup
        
               | sneak wrote:
               | That's just what they cost; MSRP is irrelevant. They're
               | not hard to find, they're just expensive.
        
             | evo_9 wrote:
             | I was using Claude 3.7 exclusively for coding, but it sure
             | seems like it got worse suddenly about 2-3 weeks back. It
             | went from writing pretty solid code I had to make only
             | minor changes to, to being completely off its rails,
             | altering files unrelated to my prompt, undoing fixes from
             | the same conversation, reinventing db access and ignoring
             | existing coding 'standards' established in the existing
             | codebase. Became so untrustworthy I finally gave OpenAi O3
             | a try and honestly, I was pretty surprised how solid it has
             | been. I've been using o3 since, and I find it generally
             | does exactly what I ask, esp if you have a well established
             | project with plenty of code for it to reference.
             | 
             | Just wondering if Claude 3.7 has seemed differently lately
             | for anyone else? Was my go to for several months, and I'm
             | no fan of OpenAI, but o3 has been rock solid.
        
               | jessmartin wrote:
               | Could be the prompt and/or tool descriptions in whatever
               | tool you are using Claude in that degraded. Have
               | definitely noticed variance across Cursor, Claude Code,
               | etc even with the exact same models.
               | 
               | Prompts + tools matter.
        
               | esskay wrote:
               | Cursor became awful over the last few weeks so it's
               | likely them, no idea what they did to their prompt but
               | its just been incredibly poor at most tasks regardless of
               | which model you pick.
        
               | sneak wrote:
               | Me too. (re: Claude; I haven't switched models.) It sucks
               | because I was happily paying >$1k/mo in usage charges and
               | then it all went south.
        
             | sneak wrote:
             | I'm firehosing about $1k/mo at Cursor on pay-as-you-go and
             | am happy to do it (it's delivering 2-10k of value each
             | month).
             | 
             | What cards are you gonna put in that chassis?
        
         | teaearlgraycold wrote:
         | What are you going to do with the LLMs you run?
        
           | chisleu wrote:
           | Currently I'm using gemini 2.5 and claude 3.7 sonnet for
           | coding tasks.
           | 
           | I'm interested in using models for code generation, but I'm
           | not expecting much in that regard.
           | 
           | I'm planning to attempt fine tuning open source models on
           | certain tool sets, especially MCP tools.
        
         | prettyblocks wrote:
         | I've been using openwebui and am pretty happy with it. Why do
         | you like lm studio more?
        
           | truemotive wrote:
           | Open WebUI can leverage the built in web server in LM Studio,
           | just FYI in case you thought it was primarily a chat
           | interface.
        
           | prophesi wrote:
           | Not OP, but with LM Studio I get a chat interface out-of-the-
           | box for local models, while with openwebui I'd need to
           | configure it to point to an OpenAI API-compatible server
           | (like LM Studio). It can also help determine which models
           | will work well with your hardware.
           | 
           | LM Studio isn't FOSS though.
           | 
           | I did enjoy hooking up OpenWebUI to Firefox's experimental AI
           | Chatbot. (browser.ml.chat.hideLocalhost to false,
           | browser.ml.chat.provider to localhost:${openwebui-port})
        
           | s1mplicissimus wrote:
           | i recently tried openwebui but it was so painful to get it to
           | run with local model. that "first run experience" of lm
           | studio is pretty fire in comparison. can't really talk about
           | actually working with it though, still waiting for the 8GB
           | download
        
             | prettyblocks wrote:
             | Interesting. I run my local llms through ollama and it's
             | zero trouble to get that working in openwebui as long as
             | the ollama server is running.
        
               | diggan wrote:
               | I think that's the thing. Compared to LM Studio, just
               | running Ollama (fiddling around with terminals) is more
               | complicated than the full E2E of chatting with LM Studio.
               | 
               | Of course, for folks used to terminals, daemons and so on
               | it makes sense from the get go, but for others it
               | seemingly doesn't, and it doesn't help that Ollama
               | refuses to communicate what people should understand
               | before trying to use it.
        
         | noman-land wrote:
         | I love LM Studio. It's a great tool. I'm waiting for another
         | generation of Macbook Pros to do as you did :).
        
         | imranq wrote:
         | I'd love to host my own LLMs but I keep getting held back from
         | the quality and affordability of Cloud LLMs. Why go local
         | unless there's private data involved?
        
           | mycall wrote:
           | Offline is another use case.
        
             | seanmcdirmid wrote:
             | Nothing like playing around with LLMs on an airplane
             | without an internet connection.
        
               | asteroidburger wrote:
               | If I can afford a seat above economy with room to
               | actually, comfortably work on a laptop, I can afford the
               | couple bucks for wifi for the flight.
        
               | seanmcdirmid wrote:
               | If you are assuming that your Hainan airlines flight has
               | wifi that isn't behind the GFW, even outside of cattle
               | class, I have some news for you...
        
               | sach1 wrote:
               | Getting around the GFW is trivially easy.
        
               | seanmcdirmid wrote:
               | ya ya, just buy a VPN, pay the yearly subscription, and
               | then have them disappear the week after you paid. Super
               | trivially frustrating.
        
               | MangoToupe wrote:
               | Woah there Mr Money, slow down with these assumptions. A
               | computer is worth the investment. But paying a cent extra
               | to airlines? Unacceptable.
        
               | diggan wrote:
               | Some of us don't have the most reliable ISPs or even
               | network infrastructure, and I say that as someone who
               | lives in Spain :) I live outside a huge metropolitan area
               | and Vodafone fiber went down twice this year, not even
               | counting the time the country's electricity grid was down
               | for like 24 hours.
        
           | PeterStuer wrote:
           | Same. For 'sovereignty ' reasons I eventually will move to
           | local processing, but for now in development/prototyping the
           | gap with hosted LLM's seems too wide.
        
           | diggan wrote:
           | There are some use cases I use LLMs for where I don't care a
           | lot about the data being private (although that's a plus) but
           | I don't want to pay XXXEUR for classifying some data and I
           | particularly don't want to worry about having to pay that
           | _again_ if I want to redo it with some changes.
           | 
           | Using local LLMs for this I don't worry about the price at
           | all, I can leave it doing three tries per "task" without
           | tripling the cost if I wanted to.
           | 
           | It's true that there is an upfront cost but way easier to get
           | over that hump than on-demand/per-token costs, at least for
           | me.
        
         | zackify wrote:
         | I love LM studio but I'd never waste 12k like that. The memory
         | bandwidth is too low trust me.
         | 
         | Get the RTX Pro 6000 for 8.5k with double the bandwidth. It
         | will be way better
        
           | marci wrote:
           | You can't run deepseek-v3/r1 on the RTX Pro 6000, not to
           | mention the upcomming 1 million context qwen models, or the
           | current qwen3-235b.
        
           | tymscar wrote:
           | Why would they pay 2/3 of the price for something with 1/5 of
           | ram?
           | 
           | The whole point of spending that much money for them is to
           | run massive models, like the full R1, which the Pro 6000 cant
        
             | zackify wrote:
             | Because waiting forever for initial prompt processing with
             | realistic number of MCP tools enabled on a prompt is going
             | to suck without the most bandwidth possible
             | 
             | And you are never going to sit around waiting for anything
             | larger than the 96+gb of ram that the RTX pro has.
             | 
             | If you're using it for background tasks and not coding it's
             | a different story
        
               | johndough wrote:
               | If the MPC tools come first in the conversation, it
               | should be technically possible to cache the activations,
               | so you do not have to recompute them each time.
        
               | pests wrote:
               | Initial prompt processing with a large static context
               | (system prompt + tools + whatever) could technically be
               | improved by checkpointing the model state and reusing for
               | future prompts. Not sure if any tools support this.
        
               | tucnak wrote:
               | https://docs.vllm.ai/projects/production-
               | stack/en/latest/tut...
        
               | storus wrote:
               | M3 Ultra GPU is around 3070-3080 for the initial token
               | processing. Not great, not terrible.
        
               | MangoToupe wrote:
               | > And you are never going to sit around waiting for
               | anything larger than the 96+gb of ram that the RTX pro
               | has.
               | 
               | Am I the only person that gives aider instructions and
               | leaves it alone for a few hours? This doesn't seem that
               | difficult to integrate into my workflow.
        
               | diggan wrote:
               | > Am I the only person that gives aider instructions and
               | leaves it alone for a few hours?
               | 
               | Probably not, but in my experience, if it takes longer
               | than 10-15 minutes it's either stuck in a loop or down
               | the wrong rabbit hole. But I don't use it for vibe coding
               | or anything "big scope" like that, but more focused
               | changes/refactors so YMMV
        
           | t1amat wrote:
           | (Replying to both siblings questioning this)
           | 
           | If the primary use case is input heavy, which is true of
           | agentic tools, there's a world where partial GPU offload with
           | many channels of DDR5 system RAM leads to an overall better
           | experience. A good GPU will process input many times faster,
           | and with good RAM you might end up with decent output speed
           | still. Seems like that would come in close to $12k?
           | 
           | And there would be no competition for models that do fit
           | entirely inside that VRAM, for example Qwen3 32B.
        
           | storus wrote:
           | RTX Pro 6000 can't do DeepSeek R1 671B Q4, you'd need 5-6 of
           | them, which makes it way more expensive. Moreover, MacStudio
           | will do it at 150W whereas Pro 6000 would start at 1500W.
        
             | diggan wrote:
             | > Moreover, MacStudio will do it at 150W whereas Pro 6000
             | would start at 1500W.
             | 
             | No, Pro 6000 pulls max 600W, not sure where you get 1500W
             | from, that's more than double the specification.
             | 
             | Besides, what is the token/second or second/token, and
             | prompt processing speed for running DeepSeek R1 671B on a
             | Mac Studio with Q4? Curious about those numbers, because I
             | have a feeling they're very far off each other.
        
           | smcleod wrote:
           | RTX is nice, but it's memory limited and requires to have a
           | full desktop machine to run it in. I'd take slower inference
           | (as long as it's not less than 15tk/s) for more memory any
           | day!
        
             | diggan wrote:
             | I'd love to see more Very-Large-Memory Mac Studio
             | benchmarks for prompt processing and inference. The few
             | benchmarks I've seem either missed to take prompt
             | processing into account, didn't share exact weights+setup
             | that were used or showed really abysmal performance.
        
         | storus wrote:
         | If the rumors about splitting CPU/GPU in new Macs are true,
         | your MacStudio will be the last one capable of running DeepSeek
         | R1 671B Q4. It looks like Apple had an accidental winner that
         | will go away with the end of unified RAM.
        
           | phren0logy wrote:
           | I have not heard this rumor. Source?
        
             | prophesi wrote:
             | I believe they're talking about the rumors by an Apple
             | supply chain analyst, Ming-Chi Kuo.
             | 
             | https://www.techspot.com/news/106159-apple-m5-silicon-
             | rumore...
        
               | diggan wrote:
               | Seems Apple is waking up to the fact that if it's too
               | easy to run weights locally, there really isn't much
               | sense to having their own remote inference endpoints, so
               | time to stop the party :)
        
         | whatevsmate wrote:
         | I did this a month ago and don't regret it one bit. I had a
         | long laundry list of ML "stuff" I wanted to play with or
         | questions to answer. There's no world in which I'm paying by
         | the request, or token, or whatever, for hacking on fun
         | projects. Keeping an eye on the meter is the opposite of having
         | fun and I have absolutely nowhere I can put a loud, hot GPU
         | (that probably has "gamer" lighting no less) in my fam's small
         | apartment.
        
       | minimaxir wrote:
       | LM Studio has quickly become the best way to run local LLMs on an
       | Apple Silicon Mac: no offense to vllm/ollama and other terminal-
       | based approaches, but LLMs have _many_ levers for tweaking output
       | and sometimes you need a UI to manage it. Now that LM Studio
       | supports MLX models, it 's one of the most efficient too.
       | 
       | I'm not bullish on MCP, but at the least this approach gives a
       | good way to experiment with it for free.
        
         | nix0n wrote:
         | LM Studio is quite good on Windows with Nvidia RTX also.
        
           | boredemployee wrote:
           | care to elaborate? i have rtx 4070 12gb vram + 64gb ram, i
           | wonder what models I can run with it. Anything useful?
        
             | nix0n wrote:
             | LM Studio's model search is pretty good at showing what
             | models will fit in your VRAM.
             | 
             | For my 16gb of VRAM, those models do not include anything
             | that's good at coding, even when I provide the API
             | documents via PDF upload (another thing that LM Studio
             | makes easy).
             | 
             | So, not really, but LM Studio at least makes it easier to
             | find that out.
        
               | boredemployee wrote:
               | ok, ty for the reply!
        
         | pzo wrote:
         | I just wish they did some facelifting of UI. Right now is too
         | colorfull for me and many different shades of similar colors. I
         | wish they copy some color pallet from google ai studio or from
         | trae or pycharm.
        
         | chisleu wrote:
         | > I'm not bullish on MCP
         | 
         | You gotta help me out. What do you see holding it back?
        
           | minimaxir wrote:
           | tl;dr the current hype around it is a solution looking for a
           | problem and at a high level, it's just a rebrand of the Tools
           | paradigm.
        
             | mhast wrote:
             | It's "Tools as a service", so it's really trying to make
             | tool calling easier to use.
        
               | ijk wrote:
               | Near as I can tell it's supposed to make _calling other
               | people 's_ tools easier. But I don't want to spin up an
               | entire server to invoke a calculator. So far it seems to
               | make _building_ my own local tools harder, unless there
               | 's some guidebook I'm missing.
        
               | xyc wrote:
               | It's a protocol that doesn't dictate how you are calling
               | the tool. You can use in-memory transport without needing
               | to spin up a server. Your tool can just be a function,
               | but with the flexibility of serving to other clients.
        
               | ijk wrote:
               | Are there any examples of that? All the documentation I
               | saw seemed to be about building an MCP server, with very
               | little about connecting an existing inference
               | infrastructure to local functions.
        
               | cchance wrote:
               | Your not spinning up a whole server lol, most MCP's can
               | be run locally, and talked to over stdio, like their just
               | apps that the LLM can call, what they talk to or do is up
               | to the MCP writer, its easier to have a MCP that
               | communicates what it can do and handles the back and
               | forth, than writing a non-standard middleware to handle
               | say calls to an API or handle using applescript, or
               | vmware or something else...
        
               | ijk wrote:
               | I wish the documentation was clearer on that point; I
               | went looking through their site and didn't see any
               | examples that weren't oversimplified REST API calls. I
               | imagine they might have updated it since then, or I
               | missed something.
        
         | zackify wrote:
         | Ollama doesn't even have a way to customize the context size
         | per model and persist it. LM studio does :)
        
           | Anaphylaxis wrote:
           | This isn't true. You can `ollama run {model}`, `/set
           | parameter num_ctx {ctx}` and then `/save`. Recommended to
           | `/save {model}:{ctx}` to persist on model update
        
             | truemotive wrote:
             | This can be done with custom Modelfiles as well, I was
             | pretty bent when I found out that 2048 was the default
             | context length.
             | 
             | https://ollama.readthedocs.io/en/modelfile/
        
             | zackify wrote:
             | As of 2 weeks back if I did this, it would reset back the
             | moment cline made an api call. But lm studio would work
             | correctly. I'll have to try again. Even confirmed cline was
             | not overriding num context
        
       | visiondude wrote:
       | LMStudio works surprisingly well on M3 Ultra 64gb and 27b models.
       | 
       | Nice to have a local option, especially for some prompts.
        
       | squanchingio wrote:
       | I'll be nice to have the MCP servers exposed like LMStudio
       | OpenAI-like endpoints.
        
       | patates wrote:
       | What models are you using on LM Studio for what task and with how
       | much memory?
       | 
       | I have a 48GB macbook pro and Gemma3 (one of the abliterated
       | ones) fits my non-code use case perfectly (generating crime
       | stories which the reader tries to guess the killer).
       | 
       | For code, I still call Google to use Gemini.
        
         | robbru wrote:
         | I've been using the Google Gemma QAT models in 4B, 12B, and 27B
         | with LM Studio with my M1 Max. https://huggingface.co/lmstudio-
         | community/gemma-3-12B-it-qat...
        
         | t1amat wrote:
         | I would recommend Qwen3 30B A3B for you. The MLX 4bit DWQ
         | quants are fantastic.
        
           | redman25 wrote:
           | Qwen is great but for creative writing I think Gemma is a
           | good choice. It has better EQ than Qwen IMO.
        
       | api wrote:
       | I wish LM Studio had a pure daemon mode. It's better than ollama
       | in a lot of ways but I'd rather be able to use BoltAI as the UI,
       | as well as use it from Zed and VSCode and aider.
       | 
       | What I like about ollama is that it provides a self-hosted AI
       | provider that can be used by a variety of things. LM Studio has
       | that too, but you have to have the whole big chonky Electron UI
       | running. Its UI is powerful but a lot less nice than e.g. BoltAI
       | for casual use.
        
         | SparkyMcUnicorn wrote:
         | There's a "headless" checkbox in settings->developer
        
           | diggan wrote:
           | Still, you need to install and run the AppImage at least once
           | to enable the "lms" cli which can later be used. Would be
           | nice with a completely GUI-less installation/use method too.
        
             | t1amat wrote:
             | The UI is the product. If you just want the engine, use
             | mlx-omni-server (for MLX) or llama-swap (for GGUF) and
             | huggingface-cli (for model downloads).
        
               | diggan wrote:
               | Those don't offer the same features as LM Studio itself
               | does, even when you don't consider the UI. If there was a
               | "LM Engine" CLI I could install, then yeah, but there
               | isn't, hence the need to run the UI once to get "the
               | engine".
        
         | rhet0rica wrote:
         | Oh, that horrible Electron UI. Under Windows it pegs a core on
         | my CPU at all times!
         | 
         | If you're just working as a single user via the OpenAI
         | protocol, you might want to consider koboldcpp. It bundles a
         | GUI launcher, then starts in text-only mode. You can also tell
         | it to just run a saved configuration, bypassing the GUI; I've
         | successfully run it as a system service on Windows using nssm.
         | 
         | https://github.com/LostRuins/koboldcpp/releases
         | 
         | Though there are a lot of roleplay-centric gimmicks in its
         | feature set, its context-shifting feature is singular. It
         | caches the intermediate state used by your last query,
         | extending it to build the next one. As a result you save on
         | generation time with large contexts, and also any conversation
         | that has been pushed out of the context window still indirectly
         | influences the current exchange.
        
           | diggan wrote:
           | > Oh, that horrible Electron UI. Under Windows it pegs a core
           | on my CPU at all times!
           | 
           | Worse I'd say, considering what people use LM Studio for, is
           | the VRAM it occupies up even when the UI and everything is
           | idle. Somehow, it's using 500MB VRAM while doing nothing,
           | while Firefox with ~60 active tabs is using 480MB. gnome-
           | shell itself also sits around 450MB and is responsible for
           | quite a bit more than LM Studio.
           | 
           | Still, LM Studio is probably the best all-in-one GUI around
           | for local LLM usage, unless you go terminal usage.
        
       | b0a04gl wrote:
       | claude going mcp over remote kinda normalised the protocol for
       | inference routing. now with lmstudio running as local mcp host,
       | you can just tunnel it (cloudflared/ngrok), drop a tiny gateway
       | script and boom your laptop basically acts like a mcp node in
       | hybrid mesh. short prompts hit qwen local, heavier ones go
       | claude. with same payload and interface we can actually get
       | multihost local inference clusters wired together by mcp
        
       | politelemon wrote:
       | The initial experience with LMStudio and MCP doesn't seem to be
       | great, I think their docs could do with a happy path demo for
       | newcomers.
       | 
       | Upon installing the first model offered is google/gemma-3-12b -
       | which in fairness is pretty decent compared to others.
       | 
       | It's not obvious how to show the right sidebar they're talking
       | about, it's the flask icon which turns into a collapse icon when
       | you click it.
       | 
       | I set the MCP up with playwright, asked it to read the top
       | headline from HN and it got stuck on an infinite loop of
       | navigating to Hacker News, but doing nothing with the output.
       | 
       | I wanted to try it out with a few other models, but figuring out
       | how to download new models isn't obvious either, it turned out to
       | be the search icon. Anyway other models didn't fare much better
       | either, some outright ignored the tools despite having the
       | capacity for 'tool use'.
        
         | t1amat wrote:
         | Gemma3 models can follow instructions but were not trained to
         | call tools, which is the backbone of MCP support. You would
         | likely have a better experience with models from the Qwen3
         | family.
        
         | cchance wrote:
         | That latter issue isnt a lmstudio issue... its a model issue,
        
         | Thews wrote:
         | Others mentioned qwen3, but which works fine with HN stories
         | for me, but the comments still trip it up and it'll start
         | thinking the comments are part of the original question after a
         | while.
         | 
         | I also tried the recent deepseek 8b distill, but it was much
         | worse for tool calling than qwen3 8b.
        
       | maxcomperatore wrote:
       | good.
        
       | v3ss0n wrote:
       | Closed source - wont touch.
        
       | xyc wrote:
       | Great to see more local AI tools supporting MCP! Recently I've
       | also added MCP support to recurse.chat. When running locally
       | (LLaMA.cpp and Ollama) it still needs to catch up in terms of
       | tool calling capabilities (for example tool call accuracy /
       | parallel tool calls) compared to the well known providers but
       | it's starting to get pretty usable.
        
         | rshemet wrote:
         | hey! we're building Cactus (https://github.com/cactus-compute),
         | effectively Ollama for smartphones.
         | 
         | I'd love to learn more about your MCP implementation. Wanna
         | chat?
        
       | zaps wrote:
       | Not to be confused with FL Studio
        
       | bbno4 wrote:
       | Is there an app that uses OpenRouter / Claude or something
       | locally but has MCP support?
        
         | eajr wrote:
         | I've been considering building this. Havent found anything yet.
        
         | cchance wrote:
         | vscode with roocode... just use the chat window :S
        
         | cedws wrote:
         | I'm looking for something like this too. Msty is my favourite
         | LLM UI (supports remote + local models) but unfortunately has
         | no MCP support. It looks like they're trying to nudge people
         | into their web SaaS offering which I have no interest in.
        
       | jtreminio wrote:
       | I've been wanting to try LM Studio but I can't figure out how to
       | use it over local network. My desktop in the living room has the
       | beefy GPU, but I want to use LM Studio from my laptop in bed.
       | 
       | Any suggestions?
        
         | skygazer wrote:
         | Use an openai compatible API client on your laptop, and LM
         | Studio on your server, and point the client to your server. LM
         | Server can serve an LLM on a desired port using the openai
         | style chat completion API. You can also install openwebui on
         | your server and connect to it via a web browser, and configure
         | it to use the LM Studio connection for its LLM.
        
         | numpad0 wrote:
         | [>_] -> [.* Settings] -> Serve on local network ( o)
         | 
         | Any OpenAI-compatible client app should work - use IP address
         | of host machine as API server address. API key can be bogus or
         | blank.
        
       | sixhobbits wrote:
       | MCP terminology is already super confusing, but this seems to
       | just introduce "MCP Host" randomly in a way that makes no sense
       | to me at all.
       | 
       | > "MCP Host": applications (like LM Studio or Claude Desktop)
       | that can connect to MCP servers, and make their resources
       | available to models.
       | 
       | I think everyone else is calling this an "MCP Client", so I'm not
       | sure why they would want to call themselves a host - makes it
       | sound like they are hosting MCP servers (definitely something
       | that people are doing, even though often the server is run on the
       | same machine as the client), when in fact they are just a client?
       | Or am I confused?
        
         | guywhocodes wrote:
         | MCP Host is terminology from the spec. It's the software that
         | makes llm calls, build prompts, interprets tool call requests
         | and performs them etc.
        
           | sixhobbits wrote:
           | So it is, I stand corrected. I googled mcp host and the
           | lmstudio link was the first result.
           | 
           | Some more discussion on the confusion here https://github.com
           | /modelcontextprotocol/modelcontextprotocol... where they
           | acknowledge that most people call it a client and that that's
           | ok unless the distinction is important.
           | 
           | I think host is a bad term for it though as it makes more
           | intuitive sense for the host to host the server and the
           | client to connect to it, especially for remote MCP servers
           | which are probably going to become the default way of using
           | them.
        
             | kreetx wrote:
             | I'm with you on the confusion, it makes no sense at all to
             | call it a host. MCP host should _host_ the MCP server (yes,
             | I know - that is yet a separate term).
             | 
             | The MCP standard seems a mess, e.g take this paragraph from
             | here[1]
             | 
             | > In the Streamable HTTP transport, the server operates as
             | an independent process that can handle multiple client
             | connections.
             | 
             | Yes, obviously, that is what servers do. Also, what is
             | "Streamable HTTP"? Comet, HTTP2, or even websockets? SSE
             | _could be_ a candidate, but it isn 't as it says
             | "Streamable HTTP" replaces SSE.
             | 
             | > This transport uses HTTP POST and GET requests.
             | 
             | Guys, POST and GET are verbs for HTTP _protocol_ , TCP is
             | the transport. I guess they could say that they use HTTP
             | protocol, which _only_ uses POST and GET verbs (if that is
             | the case).
             | 
             | > Server can optionally make use of Server-Sent Events
             | (SSE) to stream multiple server messages.
             | 
             | This would make sense if there weren't the note "This
             | replaces the HTTP+SSE transport" right below the title.
             | 
             | > This permits basic MCP servers, as well as more feature-
             | rich servers supporting streaming and server-to-client
             | notifications and requests.
             | 
             | Again, how is streaming implemented (what is "Streaming
             | HTTP")?. Also, "server-to-client .. requests"? SSE is
             | unidirectional, so those requests are happening over
             | secondary HTTP requests?
             | 
             | --
             | 
             | And then the 2.0.1 Security Warning seems like a blob of
             | words on security, no reference to maybe same-origin. Also,
             | "for local servers bind to localhost and then implement
             | proper authentication" - are both of those together ever
             | required? Is it worth it to even say that servers should
             | implement proper authentication?
             | 
             | Anyway, reading the entire documentation one might be able
             | to put a charitable version of the MCP puzzle together that
             | might actually make sense. But it does seem that it isn't
             | written by engineers, in which case I don't understand why
             | or to whom is this written for.
             | 
             | [1] https://modelcontextprotocol.io/specification/draft/bas
             | ic/tr...
        
               | diggan wrote:
               | > But it does seem that it isn't written by engineers
               | 
               | As far as I can tell, unsurprisingly, the MCP
               | specification was written with the help of LLMs, and
               | seemingly hasn't been carefully reviewed because as you
               | say, a bunch of the terms have straight up wrong
               | definitions.
        
               | kreetx wrote:
               | Using LLMs is entirely fine, but poor review for a
               | protocol definition is ..degenerate. Aren't protocols
               | supposed to be precise?
        
               | remram wrote:
               | It was written by one vendor for their own use. It is
               | miles away from an RFC or "standard"
        
         | qntty wrote:
         | It's confusing but you just have to read the official docs
         | 
         | https://modelcontextprotocol.io/specification/2025-03-26/arc...
        
       | mkagenius wrote:
       | On M1/M2/M3 Mac, you can use Apple Containers to automate[1] the
       | execution of the generated code.
       | 
       | I have one running locally with this config:                   {
       | "mcpServers": {             "coderunner": {               "url":
       | "http://coderunner.local:8222/sse"             }           }
       | }
       | 
       | 1. CodeRunner: https://github.com/BandarLabs/coderunner (I am one
       | of the authors)
        
       | smcleod wrote:
       | I really like LM Studio but their license / terms of use are very
       | hostile. You're in breach if you use it for anything work related
       | - so just be careful folks!
        
       | jmetrikat wrote:
       | great! it's very convenient to try mcp servers with local models
       | that way.
       | 
       | just added the `Add to LM Studio` button to the anytype mcp
       | server, looks nice: https://github.com/anyproto/anytype-mcp
        
       | b0dhimind wrote:
       | I wonder how LM Studio and AnythingLLM contrasts especially in
       | upcoming months... I like AnythingLLM's workflow editor. I'd like
       | something to grow into for my doc-heavy job. Don't want to be
       | installing and trying both.
        
       ___________________________________________________________________
       (page generated 2025-06-26 23:01 UTC)