🔍 Anyone here actually using Model Context Protocol (MCP)?
I’ve been exploring MCP, an open standard from @Anthropic that aims to simplify AI integrations.
In theory, this should make it easier to connect AI with databases, task managers, or even development tools. But I’m curious to know how well it actually works in practice.
If you’ve tried MCP with PostgreSQL, Todoist, or other services, I’d love to hear:
🔹 What was your experience setting it up?
🔹 What limitations have you come across?
I’ll be sharing a deeper dive soon, but if you’ve worked with MCP, drop a comment — I’d love to hear real-world experiences! 👇
You can check the full description here https://modelcontextprotocol.io/introduction if you want to dive deep into the new protocol
Replies
I found setting up local MCP servers with Claude to be a pain but the more important part is that I couldn't find a use case that actually made sense. I was sold on the idea of being able to chat with a database, Stripe, external service, etc. but the only thing that I could actually do was run a prepared set of interactions, none of which I trusted Claude to do nor would any of them yield any benefits given I still had to go do them.
Maybe I just didn't use the right servers. Is there a local database MCP for Postgres or MySQL that allows me to chat with my data, create charts, etc.? And for Stripe, is there a read-only MCP server that would let me chat with my Stripe data?
Maybe I'm just missing something but I couldn't find anything useful.
@hudsor01 My understanding is that they can only run on Claude desktop. What MCP servers are you using? Can you do interesting things with them beyond the deterministic stuff we can do with APIs? Examples?
@steveb MCP isn’t limited to Claude Desktop, I've seen it on several different clients including the copilot ecosystem. there's several different list of community servers you can use that are public. Personally I use sequential thinking, memory, server commands, Claude, software development utilities, fetch, and filesystem but with Claude's update yesterday i can remove fetch and software utilities and filesystem overlap so its overkill but I use one as a fallback. And yeah, you can do some cool stuff with MCP beyond just the basics some examples off the top of my head would be something like interactive video game storylines, you could probably get it generate content if you were an influencer and you could probably even do translations with it.
@hudsor01 Oh yes, sorry I meant only Claude can use them via the desktop app. I know Cursor and other clients allow it, too.
For Claude specifically, isn't generating storylines and social content stuff that the "normal" Claude can do? What MCP gives it superpowers to do this better?
I tried BlenderMCP with cursor.
I have no knowledge of blenders.
Modeling through natural language commands is really incredible.
Deep requirements are difficult, but personally, I think MCP will be a great tool. 👍🏻
@mocus Thanks for sharing đź’–
Fewsats
I recently did a demo on how to use MCP to buy domains and manage DNS from Cursor (it works also for any other mcp client). The experience was very smooth. We developed those libraries, happy to answer any question or share more!
https://x.com/pol_avec/status/1902390112270610721
@pol_avec Thanks for the topic, followed ^^
exclusively, set up with Claude Desktop was extremely easy and straightforward. Using sequential thinking, memory, server commands and fetch. Fetch was a game changer until today when they came out with web access.
Interesting discussion! I’m curious to hear how others are implementing the Model Context Protocol (MCP) in real-world scenarios. Has anyone noticed measurable improvements in workflow or system behavior?"
@damon_burton find out the new article on Medium https://blog.stackademic.com/model-context-protocol-mcp-in-ai-9858b5ecd9ce
Maybe you will find it usefull
Personally, setting up MCP servers in cursor has been a life saver, it makes it so much more integrated and capable than it already is. I think especially for non-technical people, being able to set up an MCP server for something like supabase or stripe would make development with AI 10x easier at least.
The best use I've had with it so far has been connecting it to browser tools, so cursor can view potential errors in the console and fix them while it is building a new feature, which saves having to re-prompt it and fix the errors. Having an MCP server for github is also very useful.
I think there is huge potential for potentially a ChatGPT like app but that integrates with common apps like notion, spotify or gmail, to make a central hub where you can, say, ask the AI to send an email, change your spotify playlist, and add a to-do task in one prompt? Still very early days and for beginners is very hard to set up.
@insightflowdata thanks for your thoughts Michael
Coda
One use case I've always wanted is to be able to type natural language to schedule simple events on my Google Calendar. Setting this up on Cursor was pretty easy. The hardest part was setting up a Google Cloud project to get the Google OAuth credentials.
Luckily, someone had created a Google Calendar MCP Server so I didn't have to figure out how to get it set up in Cursor. Having said that, if I were to get stuck, I would just ask Cursor for how to debug and get the server working. I created a tutorial on how to set up the Google Calendar MCP Server and a demo here if anyone's interested!
Hi there! I'm a product designer with no coding background who recently developed my own MCP server and Figma plugin using Cursor. Now I can analyze designs, generate code, and create components through a text interface. It's honestly amazing - like unlocking a new skill first in programming and then in Figma.