Skip to content
Agents tracked: 501 Releases (7d): 777 Agent status: no reported issuesSince last limit reset: Claude 10d · Codex 8d For AI agents: Markdown Updated Nov 21, 2026

AI security agents

Agents that attack your own systems to find exploitable weaknesses, or investigate security alerts like a SOC analyst.

By weekly package downloads, the leaders are Strix (12.4K), CAI (Cybersecurity AI) (851). By GitHub stars, the leaders are Strix (76.8K), PentAGI (34.7K), Codex Security (14.2K). Among agents with at least 20,000 downloads in the last 30 days, the fastest grower is Strix, with downloads up 21%. As of Nov 21, 2026.

Works with:
Usage
  1. 46
    BaaRai · Security

    ARTEX - AI autonomous penetration testing system (English translation; upstream: Autumn-27/ARTEX, AGPL-3.0)

    Open source
    502/4 — — 14K 10d ago 14K GitHub starsGid 50 · 2/4 signals Free (OSS)
  2. 87
    Strix · Security · open-source autonomous penetration testing agents

    An open-source multi-agent penetration testing tool that runs dynamic tests against applications, validates findings with working proofs of concept and proposes fixes; it runs locally with Docker or as a hosted service, and can run in CI/CD through GitHub Actions.

    Open sourceCLIMCPLocal modelsOwn API key
    41 12.4K up 21.1% 76.8K 11d ago 12.4K downloads a week up 21.1% 30dGid 41 Free + $29/mo
  3. 97
    OpenAI · Security · security scanning agent CLI and SDK

    Codex Security is OpenAI's application security agent; this CLI and TypeScript SDK (@openai/codex-security) scan repositories or Git changes, validate findings, generate patches and export results such as SARIF. Running scans requires Codex Security access on an OpenAI account. A separate Codex Security Cloud plugin (research preview) scans connected GitHub repositories in Codex cloud.

    Open sourceCLIAPI / SDKDesktop appOwn API key
    392/4 — — 14.2K 3d ago 14.2K GitHub starsGid 39 · 2/4 signals —
  4. 166
    VXControl · Security · open-source autonomous penetration testing agent

    A self-hosted multi-agent system that carries out penetration testing tasks in a sandboxed environment, written in Go and React and working with several model providers.

    Open sourceWeb appAPI / SDKSelf-hostedLocal modelsOwn API key
    29 — — 34.7K 47d ago 34.7K GitHub starsGid 29 Free (OSS)
  5. 231
    Alias Robotics · Security · framework for offensive and defensive security agents

    A Python framework from Alias Robotics for building and running AI security agents for reconnaissance, exploitation, privilege escalation and defensive tasks, with guardrails and human-in-the-loop controls.

    CLISelf-hostedMCPLocal modelsOwn API key
    20 851 up 27.4% 9,870 — 851 downloads a week up 27.4% 30dGid 20 Free
  6. —
    Dropzone AI · Security · SOC alert triage and investigation agent

    AI SOC analyst that investigates security alerts end to end, querying connected SIEM, EDR, cloud, identity, email and threat-intelligence tools through their APIs without moving data, and returns a verdict with supporting evidence and a report. The platform also includes an AI Threat Hunter and a chatbot for ad-hoc investigation; an AI Threat Intel Analyst is in beta. Single-tenant enterprise, multi-tenant MSSP and EU deployments are offered.

    — — — — — Website rank #1M (Tranco top-sites list) Custom
  7. —
    Microsoft · Security · security operations agents (phishing and alert triage, identity, vulnerability remediation)

    Microsoft- and partner-built agents that run in Security Copilot and inside Defender, Entra, Intune, Purview and Sentinel to automate security and IT tasks, such as the Security Alert Triage Agent (preview, extending the Phishing Triage Agent), threat intelligence briefings, threat hunting with KQL, dynamic threat detection and Conditional Access optimisation. Agents run on triggers or on demand, use Security Compute Units and come from the Security Store or custom builds.

    Web app
    — — — — — No public usage numbers Usage-based
  8. —
    Horizon3.ai · Security · autonomous penetration testing agent

    Horizon3.ai's autonomous pentesting platform that runs real attack techniques in production without installed agents, covering internal and external networks, AWS and Azure cloud, Kubernetes, Active Directory and Entra ID, web applications and phishing impact. Tests are managed in the cloud NodeZero Portal and run from a host inside the customer network; results rank fix actions and can be retested with 1-Click Verify. It offers an API, CLI and MCP server.

    Web appAPI / SDKCLIMCP
    — — — — — Website rank #150K (Tranco top-sites list) $15/mo
  9. —
    XBOW · Security · autonomous penetration testing agent

    Autonomous offensive security platform that tests web applications and APIs within a set scope, explores them as an attacker would, chains vulnerabilities into multi-step attack paths and validates each finding with a reproducible working exploit. Reports include the attack path, a decision log and remediation guidance. It runs continuously as applications change, is used through the XBOW Console and a REST API, and integrates with Jira, Sentinel and Security Copilot.

    Web appAPI / SDK
    — — — — — Website rank #183K (Tranco top-sites list) Usage-based

“#” is the position among products of the same kind (agents, frameworks, tooling) by Gid Score, our 0–100 composite of adoption, community, attention and momentum. Use the filters and sorting above, or switch to the table view. Methodology · All AI agents.

Offensive and defensive

  • Autonomous pentesting (XBOW, NodeZero, PentAGI, Strix, CAI) runs real attack techniques against applications and networks and reports what is actually exploitable, with proof.
  • SOC agents (Dropzone AI, Microsoft Security Copilot agents) investigate alerts across security tools and return a verdict with evidence.

Use responsibly

Only test systems you own or are authorised to test, in writing. Run offensive agents in isolated environments, scope them tightly and review their findings before acting.

Head-to-head comparisons

Related guides

Questions about AI security agents

Can AI do penetration testing?

Yes. Autonomous pentesting agents chain reconnaissance, exploitation and validation steps and report exploitable findings with proof. They complement rather than replace human testers, and must only be pointed at systems you are authorised to test.

Is there an open-source AI pentesting agent?

Yes. PentAGI is open-source and self-hosted; Strix is open-source and runs locally with Docker, with a hosted option; CAI publishes its source with its own licence terms.