AnPyer

Jupyter notebooks.
On Android.

AnPyer app icon AnPyer

An IPython kernel inside the app — with NumPy, pandas, SciPy and 70+ packages already installed.
No pip. No server. Everything ready when you open it.

Get it on Google Play See what it does

One-time purchase · Android 8.0+ · 64-bit ARM

AnPyer — An(droid) · Py(thon) · er — pronounced /ˈempaɪər/

Real Python notebook

A real IPython kernel. NumPy, SciPy & pandas included.

Open any .ipynb, run cells, get tables and plots. The interpreter is CPython 3.13 running inside the app — not a server in a browser tab, not a remote sandbox.

  • True IPython. Tab completion, ? help, magics, interrupt and restart.
  • 100% standard .ipynb. Files stay compatible with Jupyter and VS Code on desktop — in both directions.
  • Works offline. Your code never leaves the device to run.
AnPyer running a notebook: scipy curve_fit code, the fitted parameters and the matplotlib figure, all on one phone screen

Preinstalled

Why pip? We've got you covered.

70+ packages ship inside the app and are tested on-device — one card per area, so you can see at a glance whether your field is covered.

  • Fixed per release — there is no pip install. Check the full list first: if what you need isn't there, don't pay yet; stay tuned for the next update.
  • Missing something? Tell us. Pure-Python packages are usually a quick add.

Foundation

  • Matplotlib
  • NumPy
  • pandas
  • Plotly
  • SciPy
  • SymPy
  • requests
  • tqdm

Data analysis

  • pandas
  • Bottleneck
  • DuckDB
  • fsspec
  • Jinja2
  • missingno
  • NumExpr
  • openpyxl
  • Polars
  • PyArrow
  • python-calamine
  • seaborn
  • tabulate

Visualization

  • Matplotlib
  • Plotly
  • corner
  • missingno
  • mplfinance
  • SciencePlots
  • seaborn
  • wordcloud

Machine learning

  • scikit-learn
  • statsmodels
  • imbalanced-learn
  • LightGBM
  • SHAP

Statistics & time series

  • statsmodels
  • emcee
  • filterpy
  • lifelines
  • PyWavelets
  • scikit-posthocs
  • uncertainties

Finance

  • statsmodels
  • mplfinance
  • ta
  • tushare
  • yfinance

Science & engineering

  • SciPy
  • SymPy
  • control
  • filterpy
  • NetworkX
  • Pint
  • PyWavelets
  • uncertainties

Astronomy & geo

  • Astropy
  • corner
  • emcee
  • geopy
  • jplephem
  • pymap3d
  • pyshp
  • sgp4
  • Skyfield

Web & API

  • cryptography
  • FastAPI
  • Jinja2
  • Pydantic
  • PyYAML
  • requests
  • Starlette
  • Uvicorn

Databases & vector stores

  • asyncpg
  • Beanie
  • DuckDB
  • Elasticsearch
  • Milvus
  • pg8000
  • psycopg
  • Pydantic
  • pymongo
  • PyMySQL
  • Qdrant
  • redis
  • SQLAlchemy

Files & office

  • imageio
  • openpyxl
  • Pillow
  • pypdf
  • python-calamine
  • python-docx
  • tifffile
  • xlrd
  • XlsxWriter

Text & web scraping

  • Beautiful Soup
  • jieba
  • lxml
  • RapidFuzz
  • regex
  • requests
  • wordcloud

Your files

Process your own data. Any folder can be your workspace.

Grant access to the folder your notebook lives in, and pd.read_csv("sales_data.csv") just works. Files Python writes are synced back to that folder.

  • No broad storage permission. AnPyer uses the Android file picker and only sees the folder you choose.
  • Plain files, no lock-in. Write a notebook on your laptop, edit it on the phone, open it again on the laptop.
A notebook reading sales_data.csv from its own folder with pandas and showing describe() as a table

Optional AI assistant · bring your own key

Cells, handled by AI. Type less, do much more.

The assistant works inside your notebook, not in a side chat. It reads your cells and their outputs, edits or inserts cells — and, only if you flip the switch for the session, runs a cell, reads the result or the traceback, and fixes its own mistakes.

  • You approve each step. Every edit shows a confirmation card before it lands; every accepted change is one Undo step.
  • Your key, your provider. Any OpenAI-compatible endpoint works. Budget models do the job — our prompts and tools carry the heavy lifting.
  • It runs code only when you let it. One tap, per session. Off again when you relaunch.
The AI panel with the prompt “Plot my distance to the ISS for the next 24h” and the notebook above showing the charts it wrote and ran

You stay in the driver's seat

AI not going rogue. Hide or lock any cell.

Mark a cell hidden and the assistant gets no source, no outputs, no hash — it does not know what is there. Mark it locked and the assistant can read it but never change it.

  • Running cells is a per-session permission. Off by default, off again on relaunch.
  • Reading files in the notebook folder is a separate switch in AI settings, off by default.
  • No key configured, no network. Without a provider set up, the app makes no requests at all.
A notebook with one cell hidden from the AI and one locked; the assistant replies that the hidden cell gives it no source, no outputs and no hash

Python computes. AI expresses. The web displays.

Frontend, wait what? Py crunches, TS renders.

AnPyer's output area is a full web renderer. Crunch the numbers in one cell, ask the assistant to “plot it”, and it writes the next cell as an interactive ECharts chart — no plotting code on your side, nothing to execute: the renderer draws it instantly.

  • Three layers, each doing what it does best. Python for the math, the model for the expression, the browser for the pixels.
  • Or do it yourself. anpyer.plot takes a plain dict and renders the same way — and it's on PyPI, so the notebook still opens on desktop Jupyter.
One cell prints twelve rows of numbers; the next cell, written by the AI, renders them as an interactive ECharts line chart

Rich output

Graphs > 1000 words. Instant with AI.

Markdown, LaTeX math, Mermaid diagrams, Graphviz DOT, chemical structures (SMILES), interactive ECharts and plotly — in Markdown cells and in cell output, with the same renderer. Desktop Jupyter needs an extension for each of these; here they are built in.

  • Pinch to zoom. Any chart or diagram opens full-screen with touch gestures.
  • matplotlib as you know it. Figures render as images, CJK labels included — no tofu boxes.
A Markdown cell rendering a caffeine molecule from SMILES, a Mermaid diagram and a Graphviz DOT graph on one screen

Privacy

We run no server and collect nothing.

No analytics, no crash reporting, no telemetry SDK. Your notebooks, your files and your API key stay on your device. Nothing leaves your phone unless you configure an AI provider — and then only to that provider, straight from your device. AnPyer never downloads executable code: everything you run is either bundled with the app or written by you.

  • 0accounts to create
  • 0servers of ours
  • 0bytes of telemetry

Details in the Privacy Policy.

The honest part

Know this before you buy.

  • No pip installPackages are built in and fixed per release. If yours isn't in the list, don't pay yet.
  • matplotlib is staticFigures render as images. Interactive charts come from plotly and ECharts.
  • The assistant has limitsIt cannot touch files outside the notebook folder and cannot stop a running cell.
  • 64-bit ARM onlyarm64-v8a — which should be most phones and tablets from the last several years. Python 3.13 under the hood.

Pricing

One purchase. That's it.

$4.99one-time

No tracking. No subscription. No in-app purchases. No account. No ads. No internet needed.
You buy it. You own it. Updates included.

Get it on Google Play

The optional AI assistant needs an API key from an OpenAI-compatible provider of your choice (a pay-as-you-go API key, not a ChatGPT or Claude subscription). Usage is billed by that provider, not by us.

No tracking. No subscription. No in-app purchases. No account. No ads. No internet needed. You buy it. You own it.