Notebook-Native Web Apps
Build a Web App in Python Straight from a Jupyter Notebook
No Flask. No React. No rewrite. Turn any Python notebook into a web app — chatbots, dashboards, or internal tools — with Mercury, the open-source framework by MLJAR.
What is a Python web app?
A Python web app is any tool your code powers that other people can open in a browser — a dashboard, a chatbot, a form, or a report — without needing Python installed or knowing how to run a notebook.
Many Python web app frameworks ask you to write the app as a script or restructure the code around callbacks. Mercury works differently: write a normal Jupyter notebook, add a few widgets, and turn it into a web app without maintaining a separate frontend.
The smallest useful example
A web app in three notebook cells
Add one input to a normal notebook and use its value in the cells below. Mercury supplies the browser interface and reactive execution.
import mercury as mrname = mr.TextInput(
label="What is your name?"
)mr.Markdown(
f"## Hello {name.value}! 👋"
)Hello Ada! 👋
Change the widget and Mercury automatically recomputes the cells below it.
That’s it. Your notebook is now a web app. Try the quickstart
Apps for real data work
What can you build?
Start with the notebook you already have and shape it into the interface your users need.
AI & Chat Apps
Build chatbots, RAG tools, and AI agents with a UI — straight from your notebook.
Explore AI appsDashboards
Turn your analysis into interactive dashboards your team can actually use.
Explore dashboardsWeb Apps & Tools
Ship internal tools, calculators, and forms without touching HTML or JavaScript.
Explore web toolsA notebook-first workflow
From notebook to web app in four steps
- 01
Write a normal notebook
Use JupyterLab, MLJAR Studio, or your usual Jupyter environment. Just write Python.
- 02
Add Mercury widgets
Add a TextInput, Select, Chat, or another widget. There are no callbacks to wire up.
- 03
Preview it live
Open the live app preview in JupyterLab or MLJAR Studio and see changes as you work.
- 04
Deploy it
Self-host with Docker or publish with the managed MLJAR Cloud service.
Flexible deployment
Deploy your app, your way
Use the same Mercury notebooks with either deployment approach. Only the hosting changes.
| Deployment | Self-hosted | MLJAR Cloud |
|---|---|---|
| Setup | Your Docker host or server | Managed for you |
| Cost | Free, open-source software | Managed cloud plans |
| Control | Full infrastructure and data control | MLJAR handles infrastructure |
| Best for | Private, internal, or custom deployments | The fastest path to a shareable link |
Password protection is available in Mercury. User-based authentication is a paid option — contact MLJAR for details.
Simple by design
Why build with Mercury
No frontend code
Mercury handles layout, navigation, and styling automatically.
No callbacks
Changing a widget automatically recomputes the notebook cells below it.
Live preview
Preview the app inside JupyterLab or MLJAR Studio before deploying it.
Clean layout by default
Get a responsive interface without designing a separate frontend.
Runs from a notebook
Keep the analysis and application together in one familiar Python workflow.
Open source
Mercury is Apache-2.0 licensed and free to self-host.
Framework choices
Coming from another framework?
Here is the short version of how Mercury’s notebook-native workflow differs.
| Framework | The one-line difference |
|---|---|
| Streamlit | Streamlit reruns the app script after an interaction. Mercury works from a notebook and recomputes cells below the changed widget. |
| Gradio | Gradio is focused on quickly wrapping models and functions. Mercury supports full notebook apps, dashboards, and chat interfaces. |
| Dash | Dash uses a callback-driven application structure. Mercury keeps the workflow in a notebook without callback wiring or HTML layout code. |
See it in action
From notebook to app
Browse real Mercury apps, including chatbots, dashboards, data utilities, and custom interface examples.
FAQ
Common questions
What is a Python web app framework?+
A Python web app framework lets you build a browser-based interface — including forms, charts, dashboards, and tools — using Python. Mercury is built specifically to turn Jupyter notebooks into apps without rewriting them as a separate frontend application.
Do I need HTML, CSS, or JavaScript to build a web app in Python?+
No. Mercury generates the layout, navigation, and styling from your notebook and widgets, so you can build the application interface using Python.
Can I deploy a Mercury app for free?+
Yes. Mercury is open source under the Apache-2.0 license, and you can self-host it with Docker on your own infrastructure. MLJAR Cloud is the managed option for teams that want hosting handled for them.
What is the difference between self-hosting and MLJAR Cloud?+
Self-hosting with Docker gives you control over the infrastructure and data. MLJAR Cloud manages the application infrastructure so you can publish without configuring your own server.
Does Mercury work with JupyterLab, or only MLJAR Studio?+
It works with both. Mercury provides live app preview in JupyterLab through its extension and in MLJAR Studio. The live preview is not currently available in Google Colab or VS Code.
Is Mercury open source?+
Yes. Mercury is licensed under Apache-2.0 and developed openly on GitHub, where contributions are welcome.
Start building
$ pip install mercuryPython and a notebook are all you need to begin.