Free version of kali gpt. Works with openai and openrouter key. Openrouter key…
AI runs online
not yet reviewed by publik. It is live now; the founder has not reviewed this version yet.

Prices are for one typical use: one request of about 1,500 words sent and 375 words back. A higher quality score is better.
Kali AI Pentest has no option to run its AI on your computer.
| Model | Quality | Per 1,000 uses |
|---|---|---|
| On your computer (Ollama, 4-bit download size) | ||
| Granite 4.2 3B2.2 GB | 9 | $0 |
| Phi-4 Mini2.5 GB | 6 | $0 |
| Llama 3.1 8B4.9 GB | 7 | $0 |
| gpt-oss 20B14 GB | 9 | $0 |
| Gemma 4 31B20 GB | 19* | $0 |
| Qwen3.5 35B-A3B24 GB | 19* | $0 |
| publik API | ||
| publik-fastGLM-5.3 Flash | 42 | $1.00 |
| publik-balancedMiMo-V2.6-Pro · Kali AI Pentest | 46 | $10.00 |
| publik-smartGPT-6 Sol | 48 | $18.00 |
Quality: Artificial Analysis Intelligence Index v4.3.2, read 2026-09-22 (publik API models 2026-09-25); * = estimated by Artificial Analysis. Sizes: the Ollama library, read 2026-09-22.
Every step written out. No terminal experience needed. Pick your setup.
A Kali Linux–styled web suite that scans websites for vulnerabilities with AI — powered by your OpenAI/OpenRouter key, with a built-in engine that works with zero setup.
7 live attack engines · AI copilot · AI attack planner · built-in practice range

| 🖥️ Kali desktop UI | boot sequence, panel, dock, and a real tabbed terminal (tab-completion, history, Ctrl+C/L) |
| 🤖 AI vulnerability reports | recon evidence → severity-rated findings, risk score, remediations, next steps (MD/JSON export) |
| 🔑 BYO key + autopilot | OpenAI keys (sk-…) or OpenRouter keys (sk-or-…); OpenRouter always auto-runs the best $0 free model with live failover |
| ⚙️ Built-in engine | no key? every AI feature still works out of the box — and takes over silently if your provider fails |
| ⚔️ 7 live attack engines | nmap · sqlmap · dirb/gobuster · nikto · wpscan · searchsploit — pure TypeScript, no binaries required |
| 🧠 AI attack planner | ai plan <target> recons the host, composes a toolchain, executes step-by-step |
| 🧪 VULNSHOP range | deliberately vulnerable local shop (SQLi, auth bypass, exposed .env/.git) to test everything legally |
| 🛡️ Guardrails | authorization gate for active scans, SSRF/private-IP blocklist, read-only payloads by design |

Open the KEY VAULT tab (or ai key <provider> <key> in the terminal):
| Provider | Key format | Behavior |
|---|---|---|
| (none) | — | built-in KAI-SEC engine answers — everything works |
| OpenAI | sk-… | routed to api.openai.com (default model gpt-4o-mini, configurable) |
| OpenRouter | sk-or-… | fetches the live catalogue and always picks the best $0 free model; on 429/errors it fails over to the next free model automatically |
If a provider is unreachable, rate-limited, or the key is rejected, the suite falls back to the built-in engine and says so — requests never hang forever (60s per provider call, 150s overall budget, live KAI is thinking… Ns indicator, Ctrl+C to abort).
Self-hosting the built-in engine: it uses
z-ai-web-dev-sdk, which reads a.z-ai-configJSON file ({"baseUrl": "https://<openai-compatible-endpoint>/v1", "apiKey": "…"}) from the project dir,~, or/etc. See.z-ai-config.example. Without it, just use your own OpenAI/OpenRouter key.
Keys are stored browser-local only (localStorage) and sent per-request to the backend — never persisted server-side, never committed.
Terminal (main tab):
scan <target> [--deep] full recon + AI vulnerability report
ai ask <question> security copilot Q&A
ai plan <target> AI attack toolchain → run <n> to execute
nmap / sqlmap / dirb / gobuster / nikto / wpscan / searchsploit
range show the built-in practice target + recipes
help
AI Scanner tab: enter a target, tick the authorization box, watch the live recon pipeline, and get a formatted vulnerability report (risk score, findings, next steps) with Markdown/JSON export.

src/lib/pentest/) — no native binaries, so it runs anywhere Node runs: TCP connect scans with service/OS detection, SQLi (error/boolean/time/UNION), content discovery (500+ wordlist), web misconfig scanner, WordPress enumerator, offline exploit DBsrc/lib/ai.ts): provider chain → OpenAI → OpenRouter best-free → built-in engine, with bounded timeouts and JSON repairsrc/
├── app/api/ exec · tools · recon · pentest · range · ai/{chat,plan,analyze} · openrouter/models
├── components/kali/ boot · desktop shell · terminal · scanner · key-vault · tools-grid
└── lib/ target-guard (SSRF blocklist) · recon pipeline · pentest engines · ai · openrouter
This tool is for authorized security testing only — your own systems or targets you have written permission to assess. Active tools require a --auth acknowledgment for public targets; private/loopback ranges are blocked. Do not scan systems you do not own. The built-in VULNSHOP range exists so you can learn and demo safely.
MIT — see LICENSE.