Me too. Let’s get notebook support
me too. and just voted. It would be nice to use composer with notebook, running cell by cell, and continue from the last running cell.
This would be a killer features…
as a data scientist I agree this feature would be amazing. Ultimately, we want to take our analyses into python scripts but the notebook is key for understanding data to begin with.
this may be out of scope – but a tool that “productionizes” or at least makes a notebook “reproducible” by making it a small bit of python code would be incredible. I’m imagining completing what I can of an analysis for the day, and then asking the agent to do some work on the directory and notebook to mitigate my normal idiocy (hardcoded paths, circular cell dependencies, constants defined in cell 50, and other stuff I know I shouldn’t do but I do). it would auto-magically create that classic data-science directory:
data/
src/
notebooks/
reports/
summary.md
1 Like
Bump, I would also love Jupyter support
This would be the greatest upgrade yet to come to Cursor. Please make it happen
+1 from me.
Would love to be able to apply changes within a Jupyter Notebook, which I use a lot for debugging and small projects.
+1 Would, love support for Jupyter Notebooks
Hey everyone, I believe there is already Jupiter notebook support to some extent, but some features may not work correctly.
Can someone confirm what the current state of this looks like as I’m not a notebook user myself?
As there seems to be a lot of demand for this, I can bring up internally what features people want most in the notebooks that aren’t currently available.
Please. I see the agent does not work with jupyter notebooks yet.
You can’t apply edits to the contents of a Jupyter cell, which can be particularly tricky if the provided diff is partial and you need to carefully copy and paste over the correct part of the existing code.
3 Likes
I’d like to add my voice to supporting this one! Just the ability to apply edits to a cell or multiple cells would be huge, but being able to use cell output as context would be a step change in productivity for me and my team.
3 Likes
+1 on this, I’ve a team here who are all very keen for this feature…
1 Like
As others said, applying code doesn’t work. It would also be nice if the composer agent was able to work with Jupyter notebooks directly (I think this is not possible at least partially because of the missing apply functionality)
it’s completely not working , would love if it supports notebooks
1 Like
the current state is that neither chat nor composer can apply changes to a Jupyter notebook. so you have to copy/paste the code manually over, and this also leads to the ai being confused about the current state of the code. command K works, but I think in general all the cursor modes struggle to read Jupyter notebooks.
1 Like
It would be great if you implement Jupyter notebook support in the upcoming updates because most of the ML research is done in notebooks and only inline AI commands are working great. Composer/ Chat are not very useful due to the lack of context and apply feature!
3 Likes
+1 to this feature request! Working with Python notebook in Cursor isn’t the greatest experience comparing to other file types.
4 Likes
Would be massive for us.
Used both my votes for this!
1 Like
+1¹⁰ !! 
I use notebooks (in Cursor) as a primary IDE. Even when working with .py files, I prefer using Jupyter as a front-end to test & debug functions. There are even teams, like Nixtla, who do all dev in notebooks, & devops scripts auto-gen the py files.
At minimum, the default Cursor system prompts should be instructing composer agents to not even try updating notebooks. I’ve had to create some custom rules to make sure composer agent will only try updating .py, test with terminal, and then share code cell blocks with user, so user can paste manually. We can at leats make it aware of its limitations.
I still see some consistency issues with the custom rules, so standard system prompts would perhaps be the most effective & reliable, thus most valuable, short term solution Cursor can deliver to the data science/nb crowd until proper ipynb support is delivered. (Personally, I’d list the ipynb support as a home page feature once delivered. Should be a selling point for a significant segment of Python developers.)
A little tip for everyone too. I’ve had some successful occasions of composer agent creating a net new ipynb file, including several completed code/md cells. It just can’t update. tbh, didn’t save/don’t remember the exact prompt, but something to the effect of: asked only update/test in .py files + terminal & once complete, create a complete ipynb with validated code.
1 Like
Echoing all of this! Not sure composer works yet. Maybe this has been addressed. Thanks love the product!
2 Likes