Terminal AI for pentesters tired of browser copy-paste

What it does KaliGPT is a command-line front for several LLMs—Gemini, ChatGPT, OpenRouter, and local Ollama instances—packaged as a single kaligpt command for Linux. It wraps online search and tool-calling features into a terminal-native experience aimed at cybersecurity students and hobbyist pentesters. The project also bundles an automated installer and claims Termux support for Android.

The interesting bit The practical win is model-hopping: you can bounce between cloud APIs and an offline Ollama instance with a flag, handy when you’re on a restricted network or rationing API credits. The README advertises “agentic” behavior, though it never details which tools the agent can actually invoke or how the loop is structured, so the autonomy seems mostly aspirational right now.

Key highlights

• Supports four backends: Gemini, ChatGPT, OpenRouter, and Ollama.
• Includes online search integration and a web-chat launcher for when the terminal feels too cramped.
• Offers in-session model switching via /change-model and a /list-tools command.
• Runs on standard Linux and claims Termux compatibility for mobile workflows.

Caveats

• The authors explicitly warn the project is in active development and “don’t expect it to work flawlessly.”
• The README promises tool calling and “agentic” assistance but offers no specifics on the tool inventory or agent architecture.
• “Fine-tuned for ethical hackers” appears in the repo description, yet the README never clarifies whether that means custom weights, curated system prompts, or just themed branding.

Verdict Worth a look if you’re a student who wants quick AI answers mid-workflow inside Kali or Termux. Pass if you need a stable, transparent agent framework with documented tool integrations.