MCP Support for LLM Agents #4566
Replies: 1 comment 1 reply
-
|
Hi @moeloubani, that is very cool! As a programmer I suppose I'd want it to write generic code to solve the problem rather than directly manipulating an image, but I can imagine some people finding direct manipulation more useful. Sharp only offers a subset of libvips operations and doesn't allow streaming. Have you looked at more general purpose libvips bindings? They could potentially be more useful, though I can see they'd be much more difficult to represent for MCP. Kleis has a full libvips binding for JS here: https://github.com/kleisauke/wasm-vips There's a live playground here so you can experiment: |
Beta Was this translation helpful? Give feedback.
-
|
This was some great advice @jcupitt and I gave it a shot (I think successfully). I ended up going with a hybrid of both Sharp and the wasm package you mentioned. Thanks :) |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I created an MCP server for libvips here https://github.com/moeloubani/libvips-mcp to help me work with images inside Cursor. Would love to see if this could be added to the list of bindings or something.
It works through Sharp on Node and looks good with my testing (I'm able to use 17 different tools inside) but if anyone wants to try it out and see if they can break it I'm happy to work on that.
Anyways what it should do is let you talk naturally to your editor (like flip this image and make it grayscale) and allow the LLM to take the appropriate action with libvips.
Here is a demo of it in action: https://youtu.be/8LmZiD2gQWQ
Beta Was this translation helpful? Give feedback.
All reactions