Skip to content

Back to the catalog

Awesome Cybersecurity Agentic AI
github.comrepo

Awesome Cybersecurity Agentic AI

A curated GitHub awesome-list cataloging the emerging agentic-AI security ecosystem: dozens of Model Context Protocol servers that give LLM agents access to security tools (Nuclei, Shodan, VirusTotal, Burp Suite, Ghidra, BloodHound, x64dbg), autonomous penetration-testing and CTF-solving agent frameworks (CAI, pentagi, BreachSeek, Dark-Moon, D-CIPHER), scanners/guardrails that audit agent skills, MCP servers, and multi-agent chains for prompt injection and other risks (Aguara, Inkog, MCP Doctor, AI-Infra-Guard, ATR), agentic SOC/detection platforms (Vigil), threat-modeling frameworks for agentic systems (MAESTRO, ATFAA/SHIELD), research papers, benchmark datasets, and community/learning resources. It functions as a living index for anyone building or defending AI-agent-driven security tooling.

TL;DR A large, continuously updated awesome-list of agentic AI security tools — MCP security servers, autonomous pentest/CTF agents, agent-skill/MCP scanners, agentic SOC platforms, threat-modeling frameworks, and research/datasets.
Open original ↗
Document

What it is

A curated GitHub list (raphabot/awesome-cybersecurity-agentic-ai) split into three practical buckets the summary flattens: (1) MCP servers wrapping existing security tools so LLM agents can drive them, (2) academic research on autonomous pentest/CTF agents, and (3) tooling to secure AI agents and MCP servers themselves. The third bucket is the part worth mining if you run agents in production, not just red-team with them.

What it's good for

If you need a specific capability behind an MCP interface, the list names concrete servers with scale hints:

  • Reverse engineering / binary: x64dbg_mcp (152 tools, incl. anti-debug bypass), Bamimore-Tomi/ghidra_mcp, MCPPhalanx/binaryninja-mcp, jadx-mcp-plugin (Android decompilation).
  • Recon / OSINT: mcp-shodan, mcp-virustotal, mcp-maigret, mcp-dnstwist, urldna/mcp (phishing).
  • AD / cloud: BloodHound-MCP-AI (attack-path discovery), roadrecon_mcp_server (Azure AD), illumio-mcp-server.
  • Broad web pentest: operant-mcp (51 tools: SQLi, XSS, SSRF, IDOR, auth bypass, etc.), PortSwigger/mcp-server (Burp).

Securing the agents themselves

This is where the list carries beyond a tool index:

  • agentic-anti-patterns — catalog of 20 LLM-agent failure modes (prompt injection via tool output, MCP trust-boundary collapse, RAG poisoning, exfiltration via agent-initiated fetch, autonomy creep) with symptoms, root causes, mitigations, and detection per entry. Effectively a threat model for anyone building agents.
  • Static/dynamic scanners: Aguara (173 rules, taint tracking, rug-pull detection, offline/deterministic), Agentic Radar, AgentFence (prompt injection, secret leakage, system-prompt exposure), Tencent's AI-Infra-Guard.
  • Runtime guardrails: haldir (scoped sessions, encrypted secrets, spend limits, human-in-the-loop proxy), astracipher (DID/VC identity, PQC signatures, capability-bounded tool auth).

Caveats

It's a link list, not an evaluation — no maintenance status, no security review of the MCP servers it points to. The offensive MCP servers execute real tools with agent-level autonomy; per its own agentic-anti-patterns entry, tool output crossing back into an agent is itself a prompt-injection surface. Treat entries as leads to vet, not vetted dependencies.

Organisations named here

  • ML-DSA-65 FIPS 204
  • OWASP FinBot CTF

Technologies named here

  • nuclei-mcp
  • Illumio PCE
  • Azure Active Directory
  • illumio-mcp-server
  • Ghidra
  • aribot-mcp
  • roadrecon_mcp_server
  • x64dbg
  • Binary Ninja
  • ghidra_mcp
  • Android apps
  • x64dbg_mcp
  • mcp-dnstwist
  • BloodHound
  • Burp Suite
  • mcp-maigret
  • npm packages
  • mcp-shodan
  • mcp-virustotal
  • exposureguard-mcp
  • AI agent skills
  • haldir
  • binaryninja-mcp
  • jadx-mcp-plugin
  • BloodHound-MCP-AI
  • operant-mcp
  • MCP server
  • astracipher
  • urlDNA mcp
  • AgentFence
  • Agentic Radar
  • agentic_security
  • aguara
  • AI Agent Launch Tools
  • AI-Infra-Guard
  • AICA Agent
  • Armorer Guard
  • ATR
  • brood-box
  • BugTraceAI
  • cai
  • ClawSearch
  • ClawSec
  • DarkMoon
  • Dr.Binary
  • HOL Guard
  • Immunity Agent
  • inkog
  • Lazaretto
  • MCP Doctor
  • OpenClaw Security Suite
  • OWASP Agent Memory Guard
  • parallax
  • PentAGI
  • Pipelock
  • reaper
  • Redcells
  • Reverser Space
  • zettelforge
  • rust-in-peace
  • ShellWard
  • skillock
  • SkillPreflight
  • SourceryKit
  • Trent AI
  • Vigil
  • Vulert
  • Aeon
  • Agno
  • CrewAI
  • LangChain
  • LangGraph
  • Microsoft AutoGen
  • Microsoft Semantic Kernel
  • CyberBattleSim
  • PromptTrace
  • MAREF
  • Model Context Protocol
  • OWASP Top 10 for LLM Applications
  • EU AI Act

Techniques named here

  • STRIDE/LINDDUN threat modeling
  • SQL injection
  • Cross-site scripting
  • SSRF
  • IDOR
  • Auth bypass
  • CORS misconfiguration
  • Path traversal
  • Command injection
  • NoSQL Injection
  • Anti-debug bypass
  • DNS fuzzing
  • Prompt injection via tool output
  • MCP server trust boundary collapse
  • RAG retrieval poisoning
  • Exfiltration via agent-initiated fetch
  • Autonomy creep
  • Multi-agent vertical-domain failure
  • AI agent memory poisoning
  • Tool poisoning
  • Tool Output Mimicry
  • Jailbreaks/fuzzing attacks on LLMs
  • Secret/credential leakage
  • Rug-pull detection
  • Attack path discovery in Active Directory
  • Automated penetration testing / red teaming
  • PCAP analysis

Related in AI Security