Expose any Jupyter kernel as a lightning-fast SSH / SFTP endpoint – no browser required.

• Interactive shell – feel at home with a real Bash prompt backed by the notebook kernel.
• File transfers – full SFTP & SCP support via the Jupyter Contents API.
• Zero-config auth – ships open (no password). Plug in your own auth backend easily.
• Static host key – stable fingerprint, no more host key changed warnings.
• Warm start – pre-spawns a kernel so the first connection is instant.
• Emoji-powered UX – friendly banners & progress spinners (disable with JUPYTERSSH_UI=0).

# 1. Clone & install deps
pip install -r requirements.txt

# 2. Launch the bridge (replace URL with yours)
python juypter_ssh_server.py \
    --jupyter-url "https://my.server/lab?token=SECRET" \
    --port 2222

# 3. Connect – no password required
ssh anyuser@server -p 2222

# 4. Transfer files
scp -P 2222 data.csv anyuser@server:/workspace/

Environment variables:

• JUYPTERSSH_DEBUG=1 – verbose logging.
• JUPYTERSSH_UI=0 – disable Rich spinners.

If your notebook runs on Kaggle:

1. Open the notebook as usual.
2. In the top menu bar choose Run ▸ Kaggle Jupyter Server.
3. A dialog appears – click "VS Code Compatible URL" and copy the link (it already contains the token).
4. Pass that link to --jupyter-url when starting JuypterSSH.

Example:

python juypter_ssh_server.py --jupyter-url "https://kkb-production.jupyter-proxy.kaggle.net/k/12345abcdef/XYZTOKEN/proxy" --port 2222

PyPI release coming soon – for now install from source:

git clone https://github.com/yaelliethy/JuypterSSH.git
cd JuypterSSH
pip install -r requirements.txt

Then launch with:

python juypter_ssh_server.py --jupyter-url <YOUR_URL>

Requires Python 3.9+ and a reachable Jupyter server (local or remote).

Full technical deep-dive in Technical.md.

Topics covered:

• Component architecture
• Performance tricks (C PTY reader, batching, stdbuf)
• Security model & hardening tips
• Extending auth, shell types, file streaming

JuypterSSH defaults to:

• none auth (open access). Override SSHHandler methods to enforce keys/passwords.
• Static embedded RSA host key (2048-bit). Regenerate before production:

  ssh-keygen -t rsa -b 2048 -m PEM -C JuypterSSH -f hostkey.pem

  and paste into ssh_server.py.

Q: Does it work on Windows?
A: Server – yes (WSL recommended). Client – any OpenSSH/SFTP.

Q: Multiple users?
A: Each SSH connection gets its own Jupyter session/kernel; add auth to map users→tokens.

Q: Why not talk ZeroMQ directly?
A: Simplicity & compatibility with hosted Jupyter (Binder, Kaggle) that only expose REST/WS.

PRs & issues welcome! See Technical.md and CONTRIBUTING.md

To support this project (and my studies) please consider buying a coffee:

MIT © 2024 – JuypterSSH authors