Awesome Ai Pentesting
FreeNot checkedA curated list of AI-powered pentesting tools, frameworks, MCP servers, and resources for autonomous cybersecurity operations
About
A curated list of AI-powered pentesting tools, frameworks, MCP servers, and resources for autonomous cybersecurity operations
README
A curated list of AI-powered tools, frameworks, and resources for penetration testing, bug bounty hunting, red teaming, and cybersecurity operations.
The intersection of Artificial Intelligence and offensive/defensive security is evolving rapidly. This list tracks the most impactful open-source projects that leverage LLMs, machine learning, and autonomous agents for cybersecurity.
Maintained by SkyVanguard — Autonomous AI cybersecurity. Curated by @skyvanguard, TryHackMe Top 1% Global.
📚 Free, hands-on courses on AI pentesting, cybersecurity, AI and vibe coding — in English & Spanish → skyvanguard.com.py/courses
Stars updated: October 2026
Contents
- Autonomous AI Pentesting Agents
- MCP Servers for Security
- Bug Bounty & Recon AI
- CTF Solvers & Challenge Tools
- Vulnerability Analysis & Scanning
- Red Team & Offensive AI
- Blue Team & Defensive AI
- AI Security (Securing AI Systems)
- Fuzzing with AI/ML
- Pentest Report Generation
- OSINT with AI
- Research Papers & Resources
- Unmaintained
- Contributing
Autonomous AI Pentesting Agents
Fully autonomous agents that perform end-to-end penetration testing tasks with minimal human intervention.
| Tool | Description | AI Tool | Stars |
|---|---|---|---|
| Strix | Open-source AI hackers that run code dynamically, find vulnerabilities, and validate them through actual proof-of-concepts. | Multi-LLM (OpenAI, Anthropic, Google, OpenRouter) | |
| Shannon | Autonomous AI pentester for web apps and APIs that analyzes source code, maps attack paths, and runs real exploits: no exploit, no report. | Multi-LLM (Anthropic, OpenAI, xAI, Bedrock) | |
| PentAGI | Fully autonomous AI agents system. Docker sandboxed, 20+ built-in tools (nmap, metasploit, sqlmap), Neo4j knowledge graph. | Multi-LLM (OpenAI, Claude, DeepSeek) | |
| PentestGPT | Automated penetration testing agentic framework powered by LLMs. Published at USENIX Security 2024. | Claude Code, Codex (legacy mode: multi-LLM) | |
| CyberStrikeAI | AI-native security platform in Go combining Eino-powered agents, MCP-native tools, RAG knowledge, and human oversight in one auditable workspace. | Multi-LLM (OpenAI-compatible: OpenAI, DeepSeek, Qwen) | |
| T3MP3ST | Multi-agent offensive-security framework that turns the AI coding agent you already run into a recon-to-report hunter for web apps, CTFs, and source code, with re-derivable benchmark scores. | Claude Code, Codex, Hermes, OpenCode (or OpenRouter, Anthropic, OpenAI, xAI keys) | |
| ARTEX | Self-hosted multi-agent pentest platform: a planner chains attacks across a shared asset graph and per-task exploration graphs while workers run real tools behind a recording proxy and approval gates. | Multi-LLM (Anthropic, OpenAI, Ollama) | |
| Decepticon | Autonomous hacking agent for red teaming. Modular multi-agent architecture built on LangGraph with MCP tool integration. | Multi-LLM (via LiteLLM) | |
| Cairn | General-purpose state-space search engine validated on autonomous penetration testing: given an origin and a goal, it searches for the attack path. | Claude Code, Codex (worker backends) | |
| VulnClaw | AI pentest CLI where a model-driven agent loop runs recon, vulnerability discovery, exploitation, and reporting from natural-language input over MCP tools, with an evidence gate against hallucinated flags. | Multi-LLM (13 providers incl. OpenAI, Anthropic, DeepSeek, OpenRouter, Ollama) | |
| CyberStrike | Terminal-based offensive security harness with 13+ specialized agents, 7,600+ security skills, and 120+ OWASP test cases. | Multi-LLM (Claude, OpenAI, Gemini, Ollama, Bedrock) | |
| RedAmon | Self-hosted framework that maps the attack surface into a graph, exploits it from a Kali sandbox behind human approval gates, and opens pull requests that fix what it finds. | Multi-LLM (OpenAI, Anthropic, Bedrock, OpenRouter, Ollama) | |
| PentestAgent | AI agent framework for black-box security testing with prebuilt attack playbooks for bug bounty, red-team, and pentest. | Multi-LLM (via LiteLLM) | |
| Pentest-Swarm-AI | Autonomous pentesting with a swarm of AI agents. Orchestrates recon, classification, exploitation, and reporting specialists with ReAct reasoning. Bug bounty, monitoring, and CTF modes. | Multi-LLM (Claude, OpenAI, Gemini, Together AI, Ollama) | |
| Pentest-AI | AI pentester usable via MCP, CLI, or CI that re-runs each exploit with machine oracles before marking a finding as verified. | Claude Code, Codex, Ollama/local models | |
| Pentest Copilot | AI-powered browser-based ethical hacking assistant with agentic architecture that runs commands in the pentest environment. | Multi-LLM (OpenAI, Anthropic, Google, Mistral, Codex, Claude Code) | |
| NeuroSploit | Rust CLI pentest harness that wire-checks every finding and has a second model adjudicate it before it reaches the report. | Multi-LLM (Claude Code, Codex, Gemini, Grok) | |
| LuaN1aoAgent | Cognitive-driven autonomous AI hacker using dual-graph reasoning (causal and attack graphs) for plan-execute-reflect penetration testing. | OpenAI-compatible API | |
| Xalgorix | Self-hosted autonomous pentesting agents for real-time recon, vulnerability detection, and exploit-verified findings with a local web dashboard. | Multi-LLM (OpenAI, Anthropic, DeepSeek, Gemini, Ollama, LiteLLM) | |
| hackingBuddyGPT | Academic framework for building LLM-driven ethical hacking agents in ~50 lines of code. Research platform with reproducible privilege-escalation benchmarks. | Multi-LLM (via LiteLLM, OpenRouter by default) | |
| Darkmoon | Open source autonomous AI penetration testing platform that orchestrates 80+ offensive tools through Markdown playbooks and MCP controlled execution, covering web, cloud, Active Directory, Kubernetes and API with an evidence trail per finding. | Multi-LLM (OpenAI, Anthropic, Ollama/local) | |
| Nebula | AI-powered pentesting assistant for automating recon, note-taking, and vulnerability analysis with human decision checkpoints. | Multi-LLM (Codex, Grok, Claude, OpenRouter, Ollama, vLLM) | |
| PentestCode | Multi-agent terminal pentesting agent that runs tools, keeps persistent engagement state, and chains attacks toward a stated goal. | Multi-LLM (20+ providers incl. Claude, OpenAI, Ollama) | |
| AIRecon | Autonomous pentesting agent with a Kali Docker sandbox, Caido proxy integration, and a recon → analysis → exploit → report pipeline in a terminal UI. | OpenAI-compatible gateway (LiteLLM, vLLM, Ollama) | |
| Clearwing | Autonomous vulnerability scanner pairing a ReAct network-pentest agent (sandboxed Kali tools, human-approved exploits) with a parallel source-code hunter, built to reach Glasswing-like results with widely available models. | Multi-LLM (Anthropic, OpenAI, OpenRouter, Gemini, DeepSeek, Ollama) | |
| Cybermes | Offensive security and bug bounty agent framework with 200+ playbooks, a native Go MCP server, and a standalone CLI driven by Hermes Agent. | Hermes Agent, MCP clients (Claude, Gemini) | |
| numasec | Terminal AI security agent that drives your local tools with runbooks, keeps operation context, and tracks findings, evidence, and reports. | Multi-LLM (OpenAI, Anthropic, Google, OpenRouter, Ollama) | |
| ARTEMIS | Stanford Trinity's autonomous agent that automates vulnerability discovery with a supervisor that spawns Codex subagents. | Multi-LLM (OpenAI/OpenRouter supervisor, Codex subagents) | |
| AIDA | AI-Driven Security Assessment that connects AI to 400+ pentesting tools via MCP (Exegol container), enabling natural language-driven security workflows. | Multi-LLM (OpenAI, Claude, Ollama via MCP) | |
| Zen-AI-Pentest | Autonomous multi-agent pentest framework integrating 72+ security tools with a React dashboard, REST API, and CLI. | Multi-LLM (DeepSeek, GPT-4o, Claude, Llama) | |
| cyber-harness | Single-binary agent runtime for pentest, red team, and bug bounty that bundles scanners, proxy routing, skills, and a source-code audit distribution. | Multi-LLM (OpenAI/Anthropic-compatible, DeepSeek) | |
| REDCELL | Orchestrator and executor LLM agents run a pentest end to end inside a Kali container, steerable from an operator console, and write the report. | Multi-LLM (via LiteLLM: OpenAI, Anthropic, Google, DeepSeek, Ollama) | |
| BugTraceAI | Open-source multi-agent platform for authorized web application security testing with validation, evidence capture, and reporting. | Multi-LLM | |
| Apex | Terminal AI pentesting agent that runs blackbox and whitebox tests on web targets from a TUI or a headless CLI for CI, reporting findings with evidence and suggested fixes. | Multi-LLM (Anthropic, OpenAI, Ollama, OpenAI-compatible) | |
| vulnerator-ai | Autonomous OODA-loop pentest agent driven by Hermes, running Kali and a real browser (Playwright) over MCP; ships its own MCP front door and closes each engagement with an ATT&CK-tagged, evidence-backed report. | Multi-LLM (OpenRouter, Ollama) | |
| pwner | Agentic network and Active Directory pentest CLI that drives nmap, nuclei, impacket, netexec and kerbrute as scope-checked tool calls, and runs fully offline on a local model. | Multi-LLM (OpenAI-compatible: Ollama, Lemonade, llama.cpp, OpenAI, Anthropic) |
MCP Servers for Security
Model Context Protocol servers that bridge AI assistants with cybersecurity tools.
| Tool | Description | Stars |
|---|---|---|
| REA | Local CLI and MCP tools for agent-assisted reverse engineering of native binaries, JavaScript/Electron apps, and .NET assemblies, with evidence provenance; native deep analysis uses separately installed Hopper, Ghidra, or IDA. | |
| HexStrike AI | Advanced MCP server with 150+ cybersecurity tools. Lets AI agents autonomously run pentesting, vuln discovery, and bug bounty tools. | |
| IDA Pro MCP | AI-powered reverse engineering assistant that bridges IDA Pro with LLMs through MCP. | |
| Ghidra MCP | Ghidra MCP server with 200+ tools for AI-assisted reverse engineering, available as GUI plugin or headless server. | |
| MCP Scanner | Cisco AI Defense scanner that inspects MCP servers for threats and security findings. | |
| Eleven | Agent-first reverse engineering app where Claude, Codex, Gemini, or any ACP agent reads binaries and Android apps, renames functions, and proposes patches through Eleven's own MCP tools. | |
| JADX MCP Server | MCP server for the JADX-AI plugin enabling LLM-assisted Android APK reverse engineering. | |
| awesome-mcp-security | Curated collection of MCP security resources: tools, papers, videos, articles, and vulnerability information. | |
| APKTool MCP Server | MCP server for Apktool that lets LLMs like Claude decompile Android APKs, inspect manifests, and review the code for vulnerabilities. | |
| Google Security MCP | MCP servers that give AI clients access to Google Security Operations (Chronicle), Threat Intelligence, and other Google security products. | |
| pentestMCP | AI-powered penetration testing via MCP. 20+ security assessment tools (Nmap, Nuclei, ZAP, SQLMap) through MCP protocol. | |
| awesome-cyber-security-mcp | Curated list of MCP servers focused on cybersecurity operations. | |
| Bolt | Docker-based MCP server exposing 100+ Kali Linux security tools as plugins. Successor of mcp-for-security. | |
| mcp-server-security-standard | Open, testable security control standard (MSSS) for certifying MCP servers with levels and evidence requirements. | |
| Burp Suite MCP Server | Official PortSwigger extension that integrates Burp Suite with AI clients through MCP. | |
| mcp-vanguard | Security pentesting MCP server with 89 tools across 10 categories. Windows/WSL bridge for Kali tools integration with Claude. | |
| DomScan | Hosted domain intelligence MCP for DNS, WHOIS/RDAP, TLS, subdomains, reputation, email security, and typosquatting. |
Bug Bounty & Recon AI
AI-enhanced tools for reconnaissance and bug bounty automation.
| Tool | Description | Stars |
|---|---|---|
| Nuclei | Fast vulnerability scanner with YAML-based DSL. AI-powered template generation via -ai flag. 10k+ community templates. |
|
| Claude-BugHunter | Claude Code skill bundle for bug hunting and external red teaming: 80+ skills, 15 slash commands, and 681 disclosed-report patterns. | |
| Agentic Bug Hunter | AI bug bounty toolkit that runs recon, finding, validation, and reporting from a CLI or inside Claude Code, with free or subscription AI providers. | |
| Transilience Community Tools | Open-source Claude Code skills, agents, and slash commands for penetration testing, bug bounty hunting, and reconnaissance. | |
| HackBot | AI-powered cybersecurity chatbot using LLaMA for code analysis, scan analysis, and cybersecurity Q&A. | |
| Julius | Praetorian's LLM service fingerprinting tool that identifies Ollama, vLLM, LiteLLM, and 60+ AI servers on network endpoints. |
CTF Solvers & Challenge Tools
AI tools for solving Capture The Flag competitions and security challenges.
| Tool | Description | Stars |
|---|---|---|
| ctf-skills | Agent Skills for solving CTF challenges in web exploitation, binary pwn, crypto, reverse engineering, forensics, and OSINT, for Claude Code and other agents that support the spec. | |
| ExploitGym | Large-scale benchmark built from real-world vulnerabilities in userspace programs, V8, and the Linux kernel to evaluate AI agents at exploit development. | |
| EnIGMA | Enhanced Interactive Generative Model Agent. State-of-the-art on NYU CTF and CyBench. 3x more challenges than prior agents. (ICML 2025) | Paper |
| CTFAgent | LLM-powered CTF solver with plan-and-execute paradigm. Outperforms 88% of human CTF participants. | Paper |
| CSAW LLM CTF Attack Competition | CSAW competition using generative autonomous AI to solve CTF challenges. Major academic AI security benchmark. | Competition |
| Veria Labs CTF Agent | Autonomous CTF solver where a coordinator LLM races swarms of models on each challenge; solved 52/52 to win BSidesSF 2026. | |
| Project Muteki | Multi-agent CTF-solving swarm where a coordinator schedules several CLI agent engines toward a single goal. | |
| BoxPwnr | Framework for benchmarking LLMs and agentic strategies on HackTheBox, PortSwigger, picoCTF, TryHackMe, and other CTF platforms. | |
| Cybench | Framework for evaluating LLM agents on 40 professional-level CTF tasks from 4 competitions. | |
| LLMVault | Deliberately vulnerable, CTF-style training range for the OWASP Top 10 for LLM Applications, with scripted labs and a live mode against a local model. | |
| CVE-Bench | Benchmark for AI agents' ability to exploit real-world web application vulnerabilities. | |
| ChatGPT on CTF | Research testing whether AI-LLMs can solve CTF challenges by running commands in test environments. |
Vulnerability Analysis & Scanning
AI-powered vulnerability discovery and code analysis tools.
| Tool | Description | Stars |
|---|---|---|
| SkillSpector | NVIDIA's security scanner for AI agent skills. Detects vulnerabilities, malicious patterns, and prompt injection before a skill is installed. | |
| Semgrep | Lightweight static analysis for finding bugs and enforcing code standards, with a built-in MCP server so AI coding agents can scan code. | |
| Codex Security | OpenAI's CLI and TypeScript SDK for defining security policy and finding, validating, and fixing vulnerabilities in code. | |
| Garak | NVIDIA's LLM vulnerability scanner. Probes for hallucination, data leakage, prompt injection, toxicity, and jailbreaks. | |
| deepsec | Vercel Labs' agent-powered vulnerability scanner that runs on your own infrastructure and reviews all the code in large repos, fanning out across worker machines. | |
| DeepAudit | Multi-agent code audit platform where orchestrator, recon, analysis, and verification agents review a repository with RAG and confirm findings by running PoCs in a Docker sandbox. | |
| agent-scan | Snyk's security scanner for AI agents, MCP servers and agent skills. Detects prompt injections and tool poisoning. | |
| Mantis | Google's agent skills and ADK reference harness that threat-model a codebase, research and reproduce vulnerabilities, write patches, and report them, with Gemini or any LiteLLM-supported model. | |
| open-kritt | Self-hosted platform that orchestrates AI agents to find, de-duplicate, rank, and validate vulnerabilities in code. | |
| Agentic Security | LLM vulnerability scanner / AI red teaming kit. Protects AI systems from jailbreaks, fuzzing, and multimodal attacks. | |
| 0security | Open-source AI security agents that find vulnerabilities in code, web apps, APIs and binaries, reproduce and patch them, and run on a schedule or in CI with any model. | |
| AgentShield | Security auditor for AI agent configurations: hardcoded secrets, permission misconfigs, hook injection, and MCP server risks. | |
| LLM SAST Skills | Agent skills that turn Claude Code, Codex, and other coding assistants into a SAST scanner for web and mobile codebases. | |
| Pashov Audit Group Skills | Pashov Audit Group's AI skills for Solidity security: a pre-audit x-ray with threat model, a multi-agent smart contract auditor, and an Echidna/Medusa fuzz suite generator. | |
| AutoCVE | Agent-driven CVE discovery pipeline that selects projects, audits source code, verifies vulnerabilities, and writes reports. | |
| VulnHunter | Capital One's Claude Code skills that hunt exploitable defects forward from attacker entry points, try to disprove each finding, and propose fixes. | |
| Agentic Radar | Security scanner for agentic systems. Workflow visualization, tool identification, MCP server detection, vulnerability mapping. | |
| MEDUSA | AI-first security scanner with 40,000+ detection patterns for AI/ML apps, agents, and LLM integrations. | |
| IDAssist | IDA Pro plugin for LLM-assisted binary reverse engineering: function explanations with security analysis, rename suggestions, RAG, and a knowledge graph of the binary. | |
| Metis | Arm's agentic AI framework for deep security code review that reasons about code semantics instead of relying on hardcoded rules. | |
| OpenAnt | Knostic's LLM-based vulnerability discovery and remediation harness: one stage detects, a second attacks, and only what survives is reported. | |
| ModelScan | Scans ML models for unsafe code. Supports H5, Pickle, SavedModel, and more formats. | |
| DockSec | OWASP AI-powered Docker security scanner that explains vulnerabilities in plain English. | |
| agentgg | Agentic white-box SAST scanner whose agents follow imports and call graphs to confirm findings before reporting. CI ready. | |
| OpenHack | Agentic security scanner and verifier for codebases that exclusively uses open-source models. | |
| NVIDIA Vulnerability Analysis | NVIDIA blueprint for identifying and mitigating container security vulnerabilities with generative AI. | |
| rust-in-peace | Agent-assisted vulnerability discovery for Rust with adversarial triage, dynamic verification, and an experimental Android APK analysis profile. |
Red Team & Offensive AI
Offensive AI tools, adversarial ML, and AI red teaming frameworks.
| Tool | Description | Stars |
|---|---|---|
| Promptfoo | LLM red teaming, pentesting, and vuln scanning with 100+ attack types. Tests complete AI systems including RAG and agents. | |
| Claude-Red | Library of offensive security SKILL.md files that load on demand and turn Claude into a red team operator, from SQL injection to EDR evasion and ADCS abuse. |
|
| Adversarial Robustness Toolbox | IBM's Python library for ML Security: evasion, poisoning, extraction, inference attacks and defenses. | |
| Giskard | Python library for testing and red teaming agentic systems: scenario checks with LLM-as-judge plus a vulnerability scan for prompt injection, jailbreaks, and harmful content. | |
| PyRIT | Microsoft's Python Risk Identification Tool for red teaming generative AI systems. | |
| DeepTeam | Framework for red teaming LLMs. 10+ adversarial attacks, supports OWASP Top 10 for LLMs and NIST AI RMF. | |
| pentest-ai-agents | 50 Claude Code subagents for authorized penetration testing and offensive security research. | |
| offensive-ai-compilation | Curated list of Offensive AI resources: tools, techniques, research papers, and attack methodologies. | |
| Anamorpher | Trail of Bits tool for crafting image-scaling attacks that deliver multimodal prompt injection. | |
| EasyJailbreak | Python framework to generate adversarial jailbreak prompts against LLMs. | |
| OPFOR | Adversary emulation for AI agents and MCP servers, usable from the CLI, the IDE, or a browser extension. | |
| HackAgent | Red-team toolkit that attacks AI models and agents with jailbreaks, prompt injection, and tool misuse, scoring each reply with judge models to measure attack success rates. | |
| AgentHound | Offensive framework for agentic infrastructure: recon, credential looting, model exfiltration, and poisoning across MCP, A2A, gateways, and vector stores. | |
| RAMPART | Microsoft's pytest-native framework for safety and security testing of agentic AI applications against adversarial attacks and benign failures. | |
| Augustus | Praetorian's Go-based LLM vulnerability scanner with 210+ adversarial attacks: prompt injection, jailbreaks, encoding exploits, and data extraction. | |
| LazyOwn | Red team framework and multi-operator C2 with an LLM operator, a multi-agent hive, and 153 MCP tools for AI agents. | |
| Spikee | Reversec's modular kit for testing LLMs, guardrails, and applications against prompt injection and jailbreaks. |
Blue Team & Defensive AI
AI tools for defense, detection, and incident response.
| Tool | Description | Stars |
|---|---|---|
| AiSOC | Self-hostable AI Security Operations Center with alert fusion, LLM-agent triage, and MITRE ATT&CK investigation. | |
| Guardrails AI | Adding guardrails to LLMs. Input/output validation, content moderation, and PII detection. | |
| NeMo Guardrails | NVIDIA's toolkit for programmable guardrails on LLM systems. Content moderation, jailbreak detection. | |
| Shuffle | Open-source SOAR platform for security automation and orchestration, with AI Agent nodes inside workflows. | |
| Beelzebub | Deception framework for realistic decoys, with LLM-powered responses that sustain attacker interaction. | |
| Watcher | Thales CERT's threat detection platform that tracks CVEs, ransomware, data leaks, and suspicious domains, using Hugging Face models to summarize threats and extract IOCs. | |
| AttackGen | Incident response testing tool that uses LLMs with MITRE ATT&CK and ATLAS to generate tailored attack scenarios. | |
| numbat | Perplexity's endpoint visibility into AI agent activity, with local detection, optional pre-action blocking, and forensic reconstruction. | |
| STRIDE GPT | AI-powered threat modeling tool that generates STRIDE threat models and attack trees for an application. | |
| Vigil SOC | Open-source agentic AI SOC with Markdown playbooks and MCP integrations that runs on any supported LLM provider, including a local Ollama model. | |
| Agent Threat Rules | Open, executable detection rule format for AI agent threats, like Sigma for agents. | |
| ATHF | Agentic Threat Hunting Framework: markdown memory layer that documents hunts with the LOCK pattern so AI assistants can reuse past investigations, plus hypothesis-generation agents. |
AI Security (Securing AI Systems)
Tools for defending AI/LLM systems against prompt injection, jailbreaks, and emerging threats.
| Tool | Description | Stars |
|---|---|---|
| AI-Infra-Guard | Tencent's full-stack AI red teaming platform that scans agents, skills, MCP servers, and AI infrastructure. | |
| Superagent | SDK for AI agent safety: blocks prompt injection, redacts PII and secrets, and runs red team scenarios against agents. | |
| nono | Zero-setup sandbox that isolates AI coding agents (Claude Code, Codex, and more) with fine-grained, policy-controlled permissions. | |
| PurpleLlama | Meta's tools to assess and improve LLM security: CyberSecEval benchmarks, Llama Guard, Prompt Guard, and Code Shield. | |
| ADR | Uber's Agentic AI Detection and Response system for securing employee-facing agents like Cursor, Claude Code, and Codex. MLSys 2026. | |
| ClawSec | Prompt Security's security skills and signed advisories for OpenClaw, NanoClaw, Hermes, and Picoclaw agents: verify skills, detect configuration drift, audit environments, and gate risky installs. | |
| HOL Guard | Local-first antivirus for AI agents that reviews shell commands, file access, package installs, and MCP tool calls before they run, blocking secret exposure and prompt injection. | |
| AgentDojo | ETH Zurich's dynamic environment to evaluate prompt injection attacks and defenses for LLM agents. | |
| Pipelock | AI agent firewall that inspects HTTP, WebSocket, MCP, and A2A traffic for secret exfiltration and prompt injection. | |
| API Relay Audit | Local CLI auditing third-party AI relays and LLM proxies for prompt-injection signals, package-command text changes, error-response leakage, and Anthropic SSE anomalies, with Markdown reports. | |
| DefenseClaw | Cisco's security governance layer for AI coding agents: scans capabilities before use, inspects runtime traffic, and exports audit evidence. | |
| Arcanum PI Taxonomy | Interactive classification of prompt injection and LLM attacks for pentesters moving into AI red teaming. | |
| ZeroLeaks | CLI that red-teams LLM system prompts and agent instructions for prompt injection and extraction. | |
| Tracekit | Out-of-process signer that checks each AI agent tool call before it runs, holds risky calls for human approval, and signs every action into a tamper-evident log. | |
| SolonGate Agent Security | Local-first security gateway for AI coding agents that checks tool calls against policy before they run, with secret detection, rate limits, and a local audit trail. | |
| Open-Prompt-Injection | Toolkit and benchmark for implementing and evaluating prompt injection attacks and defenses. | |
| Awesome-LLMSecOps | LLM Security Operations in one repo. Curated links covering the full LLMSecOps lifecycle. | |
| parry-guard | Prompt injection scanner for AI coding tools (Claude Code, Codex). Runs DeBERTa/Llama transformers locally in Rust via Candle or ONNX. |
Fuzzing with AI/ML
Machine learning-guided fuzzing and automated test generation tools.
| Tool | Description | Stars |
|---|---|---|
| Buttercup | Trail of Bits' Cyber Reasoning System for DARPA AIxCC. Finds vulnerabilities with fuzzing and patches them with LLMs. | |
| FuzzyAI | CyberArk's automated LLM fuzzing tool for finding jailbreaks and security vulnerabilities in LLM APIs. | |
| EvoMaster | First open-source AI-driven tool for system-level test generation. Evolutionary algorithms for REST, GraphQL, RPC. | |
| Atlantis | Team Atlanta's Cyber Reasoning System from the DARPA AIxCC Final Competition for automated vulnerability discovery. | |
| OSS-CRS | OpenSSF framework for building and running LLM-based autonomous bug-finding and bug-fixing systems. |
Pentest Report Generation
AI tools for automated penetration testing report creation.
| Tool | Description | Stars |
|---|---|---|
| PwnDoc | Pentest report generator with customizable Docx templates and AI-assisted finding generation and QA. | |
| SysReptor | Customizable pentest reporting platform with an AI agent that drafts, reviews, and edits findings and sections. | |
| Faction | Pen test report generation and assessment collaboration platform with configurable LLM providers (OpenAI, Azure, Bedrock, Claude) for executive summaries. |
OSINT with AI
AI-powered Open Source Intelligence tools.
| Tool | Description | Stars |
|---|---|---|
| Robin | AI-powered dark web OSINT tool that uses LLMs to refine queries, filter results from dark web search engines, and summarize the investigation; runs as a web UI or MCP server. | |
| OpenOSINT | AI-powered OSINT agent with 20 tools, usable as an interactive REPL, MCP server, CLI, or web interface. | |
| Taranis AI | Advanced OSINT tool using NLP-powered news collection and AI for situational analysis. | |
| VoidAccess | Self-hosted dark web OSINT platform that fans one query out to Tor search engines and feeds, extracts entities with an LLM into a searchable graph, and exports IOCs as STIX, MISP, Sigma, or YARA. | |
| OSINTGPT | OSINT analysis tool leveraging GPT embeddings and vector search for document similarity. |
Research Papers & Resources
Papers
- PentestGPT: An LLM-empowered Automatic Penetration Testing Tool — USENIX Security 2024
- LLM Agents can Autonomously Hack Websites — GPT-4 hacks 73% of test websites
- Getting pwn'd by AI: Penetration Testing with LLMs — Systematic study
- AutoAttacker: LLM Guided Automatic Cyber-attacks — Autonomous attack systems
- EnIGMA: Enhanced Interactive Generative Model Agent for CTF — ICML 2025
- LLM Agents can Autonomously Exploit One-day Vulnerabilities — GPT-4 exploits 87% of 15 real-world one-day CVEs given the description
- Teams of LLM Agents can Exploit Zero-Day Vulnerabilities — HPTSA: hierarchical planning with task-specific agents
- Cybench: A Framework for Evaluating Cybersecurity Capabilities and Risks of Language Models — 40 professional-level CTF tasks
- Comparing AI Agents to Cybersecurity Professionals in Real-World Penetration Testing — ARTEMIS vs. 10 professionals on a ~8,000-host network
Awesome Lists
| List | Description | Stars |
|---|---|---|
| awesome-pentest | General pentesting resources. | |
| Awesome-LM-SSP | Reading list on large model safety, security, and privacy. | |
| Awesome-LLM4Cybersecurity | LLMs for cybersecurity: pentesting reasoning, exploit generation, vulnerability detection. | |
| awesome-ai-security-tools | Public-source, research, and commercial tools for AI security and AI-assisted cybersecurity. | |
| AI-Red-Teaming-Guide | Guide to adversarial testing and security evaluation of AI systems. | |
| Awesome-AI-Hacking-Agents | List of AI hacking agents. | |
| awesome-cybersecurity-agentic-ai | Autonomous security agents, frameworks, and research. | |
| awesome-MLSecOps | MLSecOps tools and resources for securing machine learning and AI systems. | |
| Awesome-Offensive-AI-Agentic-Landscape | Open-source projects, papers, benchmarks, and commercial solutions in offensive agentic AI. | |
| awesome-skills-security | Curated skill set of security-focused Claude Code skills. | |
| awesome-ai-cybersecurity | Resources for AI in cybersecurity. Collection for professionals and researchers. |
Standards & Frameworks
- OWASP AI Exchange — OWASP flagship AI security guidance (successor of the AI Security and Privacy Guide)
- MITRE ATLAS — Adversarial Threat Landscape for AI Systems
- OWASP Top 10 for LLM Applications
- OWASP Top 10 for Agentic Applications
- NIST AI Risk Management Framework
Unmaintained
Projects with no commits in the last 12 months. Still useful as reference, but expect bugs, outdated dependencies, or broken integrations.
| Tool | Description | Original Category | Last Commit | Stars |
|---|---|---|---|---|
| Awesome-GPT-Agents | GPT agents for cybersecurity: offensive, defensive, and analytical. | Awesome Lists | ||
| llm-attacks | Universal and transferable attacks on aligned language models. Academic research on adversarial attacks against LLMs. | Red Team & Offensive AI | ||
| Vulnhuntr | Zero-shot vulnerability discovery using Claude AI. Analyzes Python call chains. Found 12+ zero-days in production code. | Vulnerability Analysis & Scanning | ||
| ai-exploits | Collection of real-world AI/ML exploits for responsibly disclosed vulnerabilities. | Red Team & Offensive AI | ||
| awesome-llm-security | Tools, documents, and projects about LLM Security. | Awesome Lists | ||
| ReconAIzer | Burp Suite extension using OpenAI GPT to discover endpoints, params, URLs, and subdomains during bug bounty recon. | Bug Bounty & Recon AI | ||
| MCP-Security-Checklist | Comprehensive security checklist for MCP-based AI tools by SlowMist. Input validation, audit capability, and risk assessment. | MCP Servers for Security | ||
| prompt-injection-defenses | Every practical and proposed defense against prompt injection attacks. Comprehensive reference. | AI Security (Securing AI Systems) | ||
| AI-Goat | Learn AI security through vulnerable LLM CTF challenges. Run locally, no sign-ups or cloud fees needed. | CTF Solvers & Challenge Tools | ||
| ToolFuzz | First framework for testing LLM agent tools. Combines fuzzing with LLMs for correctness evaluation. | Fuzzing with AI/ML | ||
| PentestThinkingMCP | AI pentesting reasoning engine with Beam Search, MCTS, attack scoring for attack path planning and CTF solving. | MCP Servers for Security | ||
| LLM-Security-CTF | Learn LLM/AI security through CTF challenges: prompt injection, jailbreaks, guardrail bypass, tool abuse. | CTF Solvers & Challenge Tools |
Contributing
Contributions welcome! Please read CONTRIBUTING.md before submitting a PR.
Criteria for Inclusion
- Open-source with a GitHub repository
- Uses AI/ML in a meaningful way for security
- Actively maintained (commits within 12 months) — tools that go quiet move to Unmaintained
- Has documentation
Disclaimer
Tools listed are for educational purposes and authorized security testing only. Always obtain proper authorization before testing systems you do not own. The maintainers are not responsible for misuse.
Created by @skyvanguard
If this list helped you, consider giving it a star!
Installing Awesome Ai Pentesting
This server has no published package — it is built from source. Open the repository and follow its README.
▸ github.com/skyvanguard/awesome-ai-pentestingFAQ
Is Awesome Ai Pentesting MCP free?
Yes, Awesome Ai Pentesting MCP is free — one-click install via Unyly at no cost.
Does Awesome Ai Pentesting need an API key?
No, Awesome Ai Pentesting runs without API keys or environment variables.
Is Awesome Ai Pentesting hosted or self-hosted?
Self-hosted: the server runs locally on your machine via the install command above.
How do I install Awesome Ai Pentesting in Claude Desktop, Claude Code or Cursor?
Open Awesome Ai Pentesting on unyly.org, pick your client tab (Claude Desktop, Claude Code, Cursor) and press Install — the config is generated automatically, no JSON editing.
Related MCPs
Opencode Omniroute Plugin
OpenCode plugin for the OmniRoute AI Gateway. Drives dynamic model discovery, /connect auth flow, and multi-instance OmniRoute providers via the official @openc
by GitHub ActionsRoblox Studio
Enables AI coding tools to control Roblox Studio for workspace exploration, instance manipulation, and script management. It provides tools for playtesting, sce
by paralovAntigravity Drawio
MCP server that enables AI assistants to create, parse, render, and validate Draw.io diagrams programmatically.
by S-SUJAN-SFetch
Web content fetching and conversion for efficient LLM usage.
AWS KB Retrieval
Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.
by modelcontextprotocolSpring AI MCP Server
Provides auto-configuration for setting up an MCP server in Spring Boot applications.
llm-analysis-assistant
A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also
by xuzexin-hzMCP-Agent
A simple, composable framework to build agents using Model Context Protocol by [LastMile AI](https://www.lastmileai.dev)
by lastmile-aiSpring AI MCP Client
Provides auto-configuration for MCP client functionality in Spring Boot applications.
mcp.natoma.ai
A Hosted MCP Platform to discover, install, manage and deploy MCP servers by [Natoma Labs](https://www.natoma.ai)
Compare Awesome Ai Pentesting with
Not sure what to pick?
Find your stack in 60 seconds
Author?
Embed badge for your README
Browse similar
All ai MCPs
