A lightweight, GUI-based SSH client written in Python/PyQt5 that behaves like a minimal PuTTY clone, but adds an AI-powered command generator layer on top.
-
Pure-Python SSH
Uses Paramiko to open an interactive shell channel (xterm) and keeps the session alive in a background thread. All user commands and server responses are streamed into a scrollable QTextEdit. -
Persistent Session Log
Every line typed by the user and every byte returned by the server is appended tosession_log.jsonl(JSON Lines) with a UTC timestamp.
Example record:{"timestamp":"2025-08-05T17:43:12Z","type":"USER_COMMAND","text":"ls -l /var/log\n"} {"timestamp":"2025-08-05T17:43:12Z","type":"SERVER_OUTPUT","text":"total 876\n-rw-r--r-- 1 root root 12345 Aug 5 17:40 syslog ..."} -
AI Mode
• Activate the checkbox “Enable AI Mode”.
• Describe the task in plain English, e.g.
“find every file larger than 100 MB under /var/log”
• The assistant calls OpenAI’s Chat Completions API (gpt-3.5-turbo) with a system prompt that forces the model to return only a single, safe shell command.
• The generated command is displayed in a small preview box; the user must explicitly press “Send to Server” to execute it.
• Both the original prompt and the AI-generated command are also stored in the same JSONL file. -
Thread-Safe UI Updates
All network I/O and AI calls run in separate threads; PyQt signals keep the GUI responsive.
-
Install dependencies
pip install PyQt5 paramiko openai -
Export your OpenAI key (or edit the file)
export OPENAI_API_KEY="sk-XXX" -
Run
python main.py -
In the pop-up dialog enter the SSH password (defaults to localhost, user root; change code for other hosts).
-
Use the lower Command box for normal usage or switch to AI Mode for natural-language assistance.
• Login dialog with host, port, key-file support
• Colored terminal emulation with ANSI escape parsing
• Tabbed sessions
• Full-text search in the log viewer
• Security gate: block destructive commands (rm -rf /, dd, etc.) before execution
• Export session as HTML/PDF
• Local LLM fallback (e.g. Ollama)
MIT – do whatever you want.
A native Kotlin port lives in android/. GitHub Actions (.github/workflows/android.yml) builds it on every push that touches android/:
download the APK from the workflow run's Artifacts, or push a v* tag to publish it on the Releases page.
Enter host, user, password and (optionally) your OpenAI key on the login screen; tick AI Mode to turn a description into a command, then tap the suggestion to run it.