Deepnote is a cloud data workspace built around collaborative notebooks, now positioning itself for both humans and agents. It supports Python and SQL in the same notebook, offers data apps and dashboards, ETL pipelines, ML model training and serving, over 100 integrations with warehouses and tools, GPU access, scheduling, versioning and code review, plus Deepnote AI for generating insights and querying data in natural language. The thing it does better than anything else is collaboration. Jupyter has been the default notebook for over a decade and it has never solved multiple people working in the same notebook at once.
The standard workflow is still exporting, emailing, committing giant JSON diffs to git and resolving merge conflicts in serialised cell outputs, which is as unpleasant as it sounds. Deepnote gives you Google Docs-style simultaneous editing with comments, and once you have worked that way on a shared analysis you resent going back. For teaching in particular this is transformative: an instructor can open a student's notebook and fix the problem in place rather than exchanging screenshots. The education footprint backs that up, with 96 of the top 100 universities using it, which means a lot of students already have institutional access and should check before paying.
Python and SQL in one notebook is the other underrated feature. The ordinary analysis loop involves querying a warehouse, pulling results into pandas, doing something, then going back for more data, and doing that across two tools with an export in between is a constant small tax. Removing it changes how fluidly you work. Deepnote AI is competent rather than remarkable.
Natural language querying, automated insight generation, agent-based automation for things like fraud detection. It is useful, it will save you time on boilerplate, and it is broadly comparable to what every notebook and IDE now ships. I would not choose the platform for the AI. I would choose it for the collaboration and treat the AI as a bonus.
Pricing is where I have reservations. Free gives you 3 editors, 5 projects, limited AI, unlimited basic machines at 5GB RAM and 2 vCPU, and 7 day revision history. That is a fair free tier and the unlimited basic machines are more generous than metered compute. Team is $39 per editor per month billed yearly with unlimited editors and projects, full AI, background execution, scheduled notebooks, 30 day history and monthly credits covering AI, CPU and GPU usage.
For a team of eight that is over $3,700 a year, and you are paying for a notebook environment whose local equivalent costs nothing. The value has to come from collaboration and managed infrastructure, and for some teams it genuinely does. For a solo analyst it plainly does not. Cloud-only is a hard constraint for anyone with data residency or confidentiality requirements, and no amount of SOC 2 compliance resolves an organisational policy that data does not leave the building.
One market note. Search interest has declined over the past year in a category that now includes Google Colab, Hex, Databricks notebooks, VS Code's notebook support and Jupyter itself. That is not a crisis, and it is a reason to weigh how much of your workflow you build around any one hosted platform. Three point six.
Excellent at the specific thing it is best at, priced for teams rather than individuals, in a market that is getting more crowded.