AI for IT & Cybersecurity
63 tools. Prices and free tiers shown only where we checked them on the official site.
Cert study and labs: TryHackMe's Echo tutor and HTB Coach. Scripting and PowerShell: practice with GitHub Copilot or Warp, then scan your own scripts for free with Semgrep or Snyk, or try OpenAI's Codex Security CLI on a practice repo you own. Before you write any fixes, have your coding agent draft a threat model of the same repo with Grim Prompter and check each finding against the code it cites. SOC and malware practice: stand up Wazuh, the free open-source SIEM, or Security Onion for network monitoring and threat hunting, in your home lab and point it at a few VMs, paste a suspicious PowerShell sample into VirusTotal Code Insight, let an AI assistant walk you through a compiled sample in Ghidra with GhidraMCP or REA (isolated VM only, and only software you're allowed to analyze) and build a small alert workflow on the Tines free edition. Help desk and MSP work: start the 30-day free trial of Atera, the RMM and ticketing platform many MSPs run, and practice working tickets and generating scripts with its built-in AI Copilot. For internal IT help desk roles, the Freshservice free trial lets you practice incidents, a knowledge base and a self-service portal, and the free plan of Jira Service Management (up to 3 agents, no time limit) lets you build a practice service desk with a portal, queues and SLAs that you can show in interviews. Zendesk shows up in a lot of help desk job postings, and its 14-day free trial is enough to learn the ticket workflow. Many larger employers run ServiceNow, and its free Personal Developer Instance gives you an admin sandbox to learn incidents, requests and the service catalog. Vulnerability scanning: learn how analysts decide what to patch first with CVE Intel: scan the apps on your own PC or a home-lab VM, then compare each CVE's risk score with its CISA KEV and EPSS evidence. Then run Nuclei against your own lab VMs (never anything you don't own) and try its -ai option to turn a plain-English check into a template. For the bigger picture, self-host CVEngine on a spare Linux VM and point it at a lab range or a domain you own: it chains subdomain discovery, port scans, nuclei and exploited-CVE matching, and every validation check records the exact command it ran, so you can learn the commands behind each finding. To see how AI pentesting agents chain and prove attacks, run Strix or Shannon against a deliberately vulnerable practice app in an isolated lab, never against production or anything you aren't authorized to test. PentAGI is a self-hosted take on the same idea with a web UI: its agents run nmap, Metasploit and sqlmap in a Docker sandbox and can use local models through Ollama, so a home lab with a vulnerable VM is a good place to watch a full recon-to-report run. For code review on a practice repo you own, try 0 from 0sec Labs, a free open-source harness with a local web console, CLI and MCP server that uses your own model provider; its default shell runs on your host, so give it an isolated executor or a throwaway VM and check each finding against the code before you fix anything. AI security: before you install an MCP server, agent skill or browser extension, look it up in Neo's free Neoverse catalog (email sign-in) and read the verdict and key finding, then still check what the tool can access yourself. Point garak at a local model to see how prompt injection and jailbreaks actually work, then write your own red-team tests with promptfoo. Before you install a skill or MCP server for your AI coding agent, scan it with NVIDIA's free SkillSpector; its --no-llm static scan works offline and flags prompt injection, data exfiltration and risky code. Then run Snyk Agent Scan to inventory every MCP server and skill your agents already have installed, and if you publish your own skills, add Cisco Skill Scanner as a pre-commit hook or GitHub Action so each change gets a SARIF report. On an Apple silicon Mac, go one step further and run your agent inside Strands Box, which blocks anything your policy doesn't permit. You can also put the free HOL Guard in front of the coding agent you already use so risky commands, secret reads and MCP calls wait for your approval. To practice the detection side, install Cisco's open-source Project Quarry plugin and write a rule that flags your own agent reading a secret and then sending data out, the same skill SOC analysts use when they write detection rules for logs. If you build agents, BotGauge shows how red-teaming an agent works: attack it, replay the failure and turn it into a regression test. When an agent needs to sign in to a site for you, use 1Password for Agents so you approve each login and the agent never sees the password. If you're learning Kubernetes for a DevOps or cloud support role, install the free Radar against a kind or minikube cluster in your home lab, break a deployment on purpose and compare what the UI shows with what your AI agent finds through Radar's read-only MCP mode. To see how AI agents hunt bugs in code, run Google's Mantis on a deliberately vulnerable practice app inside an isolated VM. To see what an AI SOC looks like from the inside, run the free, open-source AiSOC in your home lab, load its sample data and read the Investigation Ledger to check each AI triage verdict against the evidence, the way a Tier 1 analyst reviews alerts. If you're heading toward pentesting, point RedAmon only at a deliberately vulnerable practice app in an isolated lab and study each recon and exploit step it takes. To learn AI red teaming, use Microsoft's free PyRIT against a local model you run yourself and compare which jailbreak and prompt-injection attempts it scores as successful. For endpoint visibility practice, run Arthur's free Deckard quickstart on your own machine: it uses osquery to find the AI agents installed there and shows them in a local Kibana, the same inventory-to-SIEM flow a SOC uses. If you run coding agents yourself, Failproof AI logs every run and blocks risky tool calls, and its local audit shows what your agents actually did. To check whether your own agent has already leaked a key, run Stroq's npx @stroq/cli sent on your own machine: it compares past Claude Code, Codex and Cursor sessions with your credential files and names any key that went out, the same exposure review an incident responder does. To keep an agent away from your real files, run devvm in a project folder (macOS or Linux): it gives that project its own VM that mounts only the project folder and starts on an allowlist network. On a Windows PC with Pro or Enterprise, sandbox-mcp boots a throwaway Windows 11 machine so you can test an installer or app on a clean PC and delete it afterward. To see AI security testing from the attacker's side, run Tencent's free AI-Infra-Guard in Docker on your lab network (never exposed to the internet) and scan your own Ollama box or an MCP server you wrote. GRC and compliance: if risk or compliance analyst roles interest you, self-host the free Community edition of CISO Assistant and run a small ISO 27001 or NIST CSF assessment of your own home lab; it's a portfolio piece hiring managers can follow, and its optional AI chat can run on a local model through Ollama. Home lab: run local models with Ollama (and give them richer answers in Open WebUI with Frontier UI), or ship on-device AI inside a small desktop app with Microsoft's Foundry Local SDKs, knowing that big security-focused open models like Mistral Large 4 need data-center GPUs, so try them through the API or OpenRouter instead and turn notes into quizzes with Gemini Notebook. The SOC copilots further down (Security Copilot, Purple AI, Charlotte AI) are enterprise tools you'll likely meet on the job, so their docs are worth a read even before you can use them.
No tools match. Try a different word or switch the filter to All.