An AI red team that piggybacks on your LLM bill

What it does

CyberStrike is a terminal-based offensive-security agent that plugs into your existing LLM subscription—Claude, GPT, Gemini, or a local Ollama instance—and runs autonomous penetration tests from a TUI. It handles reconnaissance, vulnerability discovery, exploitation, and reporting across web applications, mobile, cloud, and internal networks. A built-in Chromium browser called HackBrowser captures traffic manually or autonomously and routes it through eight specialized proxy testers for flaws like IDOR, injection, and business-logic bypasses.

The interesting bit

Rather than wrapping a single API, CyberStrike acts as an intelligence layer that normalizes structured outputs across 15+ providers, guards context between test phases, and orchestrates multi-step attack chains. Its proxy testers follow a three-gate confirmation protocol—baseline request, attack, response comparison—so a finding is only reported when there is a measurable, reproducible difference.

Key highlights

• Supports 15+ LLM providers, including fully offline operation via Ollama or LM Studio
• 13+ specialized agents covering OWASP web testing, mobile (Frida/MASTG), cloud (CIS benchmarks), and internal networks
• HackBrowser performs role-aware crawling and feeds captured traffic to eight parallel sub-testers automatically
• Bolt enables remote tool execution across multiple servers paired with Ed25519 keys
• AGPL-3.0 license

Verdict

Grab it if you already pay for an LLM and want a structured, terminal-driven red-team assistant. Look elsewhere if you want a standalone security framework with no AI dependencies.