# AlpacaX

> AlpacaX makes Alpacon—AI-native privileged access management (PAM) for
> AI agents, engineers, and CI/CD pipelines. Other tools give agents access;
> Alpacon judges every command an agent runs against the session's declared
> purpose—before it reaches the host—and leaves one queryable audit record
> per session.

## Product

- [Alpacon Overview](https://www.alpacax.com/alpacon/): One control plane for every path to production—browser, CLI, API, and MCP—under the same access rules, the same judgment, and the same record
- [Features](https://www.alpacax.com/alpacon/features/): Access, interfaces, execution control, and auditing—outbound-only zero-trust tunnels, runtime command judgment, Slack approval, session recording, RBAC, SAML SSO
- [How It Works](https://www.alpacax.com/alpacon/how-it-works/): Outbound-only tunnel architecture and the three layers—access, execution, audit
- [Pricing](https://www.alpacax.com/pricing/): Per-server pricing, SaaS or self-hosted, with pilots available—request a quote sized to your fleet
- [Extensions](https://www.alpacax.com/alpacon/extensions/): IP management, DNS, and private CA—Enterprise add-on
- [Security](https://www.alpacax.com/security/): What listens on your network, what happens to secrets, what gets judged, and what gets recorded—the security review's first questions, answered

## Solutions

- [AI Agent Operations](https://www.alpacax.com/solutions/ai-agent-ops/): Scoped production access for Claude, Codex, and SRE bots, with every command judged before it runs
- [Team Access](https://www.alpacax.com/solutions/team-access/): Hundreds of servers, one login, nothing to install—terminal, files, editor, and tunnels in the browser
- [CI/CD & Automation](https://www.alpacax.com/solutions/automation/): Pipelines as identities—deny-by-default allowlists, IP and time bounds, rotation policies
- [Vendor Access Management](https://www.alpacax.com/solutions/vendor-access-management/): Time-boxed, recorded third-party access that expires with the session

## Developer Tools

- [Documentation](https://docs.alpacax.com/): Setup guides, API reference, MCP integration, CLI usage
- [CLI Reference](https://docs.alpacax.com/reference/cli/): Terminal-based server access, transfers, and batch commands under the same policy
- [MCP Integration](https://www.alpacax.com/alpacon/features/): Model Context Protocol interface for Claude, Codex, and VS Code—agents run through the same access rules, the same judgment, and the same record

## Integrations

- Identity: SAML SSO, LDAP, MFA (biometrics, hardware keys)
- CI/CD: GitHub Actions, Docker image, REST API service tokens
- Approval and alerts: Slack, email, browser push, webhooks (Discord, Microsoft Teams, Telegram)

## Content

- [Blog](https://www.alpacax.com/blog/): Engineering deep dives, incident analyses, and product updates on AI agent execution control
- [Newsroom](https://www.alpacax.com/newsroom/): Company news, events, and milestones
- [How we patched php2shell in six minutes with an AI agent](https://www.alpacax.com/blog/how-we-patched-php2shell-in-six-minutes-with-an-ai-agent/)
- [The model you approved isn't the model running today](https://www.alpacax.com/blog/the-model-you-approved-isnt-the-model-running-today/)
- [Loopjacking: what you approved isn't what ran](https://www.alpacax.com/blog/loopjacking-what-you-approved-isnt-what-ran/)
- [Product update: how a request arrived, and what reaches you](https://www.alpacax.com/blog/product-update-how-a-request-arrived-and-what-reaches-you/)
- [What is agentic AI? The definition that actually holds up](https://www.alpacax.com/blog/what-is-agentic-ai-the-definition-that-actually-holds-up/)
- [Inside the AI agent mass exploitation of PaperCut NG/MF](https://www.alpacax.com/blog/inside-the-ai-agent-mass-exploitation-of-papercut-ng-mf/)
- [Two CrowdStrike failures, one kernel-mode security risk](https://www.alpacax.com/blog/two-crowdstrike-failures-one-kernel-mode-security-risk/)
- [MCP security can't stop at the install review](https://www.alpacax.com/blog/mcp-security-cant-stop-at-the-install-review/)
- [What AI agent access control actually has to check](https://www.alpacax.com/blog/what-ai-agent-access-control-actually-has-to-check/)
- [Stay on Supabase and Vercel, or run your own servers?](https://www.alpacax.com/blog/supabase-and-vercel-or-your-own-servers/)
- [Inside JadePuffer, the first documented AI agent ransomware attack](https://www.alpacax.com/blog/inside-jadepuffer-the-first-documented-ai-agent-ransomware-attack/)
- [How BNOW fixes production without a DevOps team](https://www.alpacax.com/blog/bnow-fixing-production-without-a-devops-team/)
- [A human caught the AI agent. That's not a control you can rely on](https://www.alpacax.com/blog/a-human-caught-the-ai-agent-thats-not-a-control-you-can-rely-on/)
- [AI agent framework harvested thousands of credentials in under 6 hours](https://www.alpacax.com/blog/ai-agent-framework-harvested-thousands-of-credentials-in-under-6-hours/)
- [Bank regulators wrote AI agents out of model-risk review](https://www.alpacax.com/blog/bank-regulators-wrote-ai-agents-out-of-model-risk-review/)
- [MCP least privilege has two layers, not one](https://www.alpacax.com/blog/mcp-least-privilege-has-two-layers-not-one/)
- [Product update: what gets re-checked, and what gets written down](https://www.alpacax.com/blog/product-update-what-gets-re-checked-and-what-gets-written-down/)
- [A token that can reach your API isn't cleared to run any command](https://www.alpacax.com/blog/a-token-that-can-reach-your-api-isnt-cleared-to-run-any-command/)
- [Zero standing privilege: a 7-question audit for your stack](https://www.alpacax.com/blog/zero-standing-privilege-a-7-question-audit-for-your-stack/)
- [MFA fatigue is a scoping problem, not a user problem](https://www.alpacax.com/blog/mfa-fatigue-is-a-scoping-problem-not-a-user-problem/)
- [AI compliance software doesn't solve contractor AI agent risk](https://www.alpacax.com/blog/ai-compliance-software-doesnt-solve-contractor-ai-agent-risk/)
- [A CVSS 10.0 bug turned Metabase's password reset into full admin](https://www.alpacax.com/blog/a-cvss-10-0-bug-turned-metabases-password-reset-into-full-admin/)
- [Identity mapping in PAM: four handoffs, one fails silently](https://www.alpacax.com/blog/identity-mapping-in-pam-four-handoffs-one-fails-silently/)
- [AI vendor questionnaire checklist: 6 runtime questions to score](https://www.alpacax.com/blog/ai-vendor-questionnaire-checklist-6-runtime-questions-to-score/)
- [AI agent governance means more than access control](https://www.alpacax.com/blog/ai-agent-governance-means-more-than-access-control/)
- [Verification guidance exists. Turning it into a rule is still your job.](https://www.alpacax.com/blog/verification-guidance-exists-turning-it-into-a-rule-is-still-your-job/)
- [Product update: what gets decided before a command runs](https://www.alpacax.com/blog/product-update-what-gets-decided-before-a-command-runs/)
- [Why AI governance pilots don't survive scale](https://www.alpacax.com/blog/why-ai-governance-pilots-dont-survive-scale/)
- [Out-of-band approval: why an agent's own channel can never be the one that approves](https://www.alpacax.com/blog/out-of-band-approval-why-an-agents-own-channel-cant-approve-itself/)
- [HIPAA day-90 audit: what session expiry misses](https://www.alpacax.com/blog/hipaa-day-90-audit-what-session-expiry-misses/)
- [Kubernetes access control governs access, not commands](https://www.alpacax.com/blog/kubernetes-access-control-governs-access-not-commands/)
- [Session-scoped sudo: bind OS-level root to a session, not a sudoers file](https://www.alpacax.com/blog/session-scoped-sudo-bind-os-level-root-to-a-session-not-a-sudoers-file/)
- [EU AI Act high-risk deadline moved to 2027. What didn't?](https://www.alpacax.com/blog/eu-ai-act-high-risk-deadline-moved-to-2027-what-didnt/)
- [Two 2026 CVEs turned security tools into the attack path](https://www.alpacax.com/blog/two-2026-cves-turned-security-tools-into-the-attack-path/)
- [Product update: what the audit log stops keeping—and stops missing](https://www.alpacax.com/blog/product-update-what-the-audit-log-stops-keeping-and-stops-missing/)
- [SSO and MFA verify the login, not what the session does next](https://www.alpacax.com/blog/session-layer-catches-what-sso-misses/)
- [Credential theft: entitlements without purpose](https://www.alpacax.com/blog/credential-theft-entitlements-without-purpose/)
- [Command filtering fails two ways: GTFOBins, and a shell that rewrites what you matched](https://www.alpacax.com/blog/command-filtering-fails-two-ways-gtfobins-and-a-shell-that-rewrites-what-you-matched/)
- [Offboarding access revocation: what a departing engineer leaves running](https://www.alpacax.com/blog/offboarding-access-revocation-what-a-departing-engineer-leaves-running/)
- [AI agent execution control: what it is, and why AI-native PAM needs it](https://www.alpacax.com/blog/ai-agent-execution-control-what-it-is-and-why-ai-native-pam-needs-it/)
- [AgentForger: how one link forged a rogue AI agent with a borrowed employee session](https://www.alpacax.com/blog/agentforger-how-one-link-forged-a-rogue-ai-agent-with-a-borrowed-employee-session/)
- [AI vendor risk assessment in 2026: the runtime questions your questionnaire is missing](https://www.alpacax.com/blog/ai-vendor-risk-assessment-the-runtime-questions-your-questionnaire-is-missing/)
- [ISO 42001 Stage 1 checklist: the documentation to have ready before your audit](https://www.alpacax.com/blog/iso-42001-stage-1-checklist-the-documentation-to-have-ready-before-your-audit/)
- [Zero standing privilege by design: make it a property of the session, not the token](https://www.alpacax.com/blog/zero-standing-privilege-by-design-a-property-of-the-session-not-the-token/)
- [ISO 42001 audit checklist: the controls a policy document can't satisfy](https://www.alpacax.com/blog/iso-42001-audit-checklist-the-controls-a-policy-document-cant-satisfy/)
- [Alignment isn't enough: containing a misaligned agent's actions takes runtime execution control](https://www.alpacax.com/blog/alignment-isnt-enough-containing-a-misaligned-agent-takes-runtime-execution-control/)
- [AI governance on paper vs. governance during the task](https://www.alpacax.com/blog/ai-governance-on-paper-vs-governance-during-the-task/)
- [Product update: the judgment behind every approval](https://www.alpacax.com/blog/product-update-the-judgment-behind-every-approval/)
- [An autonomous AI agent breached Hugging Face—here's the kill chain, and where execution control bounds it](https://www.alpacax.com/blog/an-autonomous-ai-agent-breached-hugging-face-heres-the-kill-chain/)
- [LegacyHive: a Windows zero-day you can't patch your way out of](https://www.alpacax.com/blog/legacyhive-a-windows-zero-day-you-cant-patch-your-way-out-of/)
- [AutoJack: one webpage, one MCP socket, host-level access](https://www.alpacax.com/blog/autojack-one-webpage-one-mcp-socket-host-level-access/)
- [A command log won't tell you what your agent actually did.](https://www.alpacax.com/blog/a-command-log-wont-tell-you-what-your-agent-actually-did/)
- [JIT access control is for humans. What should we have for AI agents?](https://www.alpacax.com/blog/jit-access-control-is-for-humans-what-should-we-have-for-ai-agents/)
- [Adding AI agents shouldn't mean adding more security exposure](https://www.alpacax.com/blog/adding-ai-agents-shouldnt-mean-adding-more-security-exposure/)
- [The gate worked. The database was still dumped.](https://www.alpacax.com/blog/the-gate-worked-the-database-was-still-dumped/)
- [Work Sessions: the audit unit your CISO actually needs](https://www.alpacax.com/blog/work-sessions-the-audit-unit-your-ciso-needs/)
- [Agentic AI security is a stack, not a single control](https://www.alpacax.com/newsroom/archive/agentic-ai-security-is-a-stack-not-a-single-control/)

## Company

- [About](https://www.alpacax.com/about/): Team and mission; founded 2022; offices in Menlo Park, Seoul, and Daejeon; backed by Rabbit Ventures and KAIST Ventures; partner programmes include POSCO IMP
- [Careers](https://www.alpacax.com/careers/): Open roles at AlpacaX

## Trust & Security

- [Trust Center](https://trust.alpacax.com/): SOC 2 Type 2 (in progress), live control status, reports, subprocessors
- [Security Overview](https://www.alpacax.com/security/): Architecture, judgment, disclosure policy, and availability

## Optional

- [GitHub](https://github.com/alpacax): Open source—alpamon, alpacon-cli, alpacon-mcp
- [App](https://alpacon.io/): Sign in to an existing Alpacon workspace

## 한국어 콘텐츠 (Korean)

### 제품

- [Alpacon 소개](https://www.alpacax.com/kr/alpacon/): 브라우저, CLI, API, MCP까지 프로덕션으로 가는 모든 경로를 하나의 접근 규칙과 판정, 기록으로 통제하는 컨트롤 플레인
- [주요 기능](https://www.alpacax.com/kr/alpacon/features/): 접근, 인터페이스, 실행 제어, 감사에 걸친 아웃바운드 전용 제로 트러스트 터널, 실시간 명령 판정, Slack 승인, 세션 녹화, RBAC, SAML SSO
- [작동 원리](https://www.alpacax.com/kr/alpacon/how-it-works/): 아웃바운드 전용 터널 아키텍처와 접근·실행·감사 3계층
- [요금](https://www.alpacax.com/kr/pricing/): 서버 수 기준 요금, SaaS 또는 자체 호스팅, 파일럿 가능. 서버 규모에 맞춘 견적 요청
- [Extensions](https://www.alpacax.com/kr/alpacon/extensions/): IP 관리, DNS, Private CA를 담은 Enterprise 전용 애드온
- [보안](https://www.alpacax.com/kr/security/): 네트워크에서 무엇이 수신 대기 중인지, 시크릿은 어떻게 처리되는지, 무엇이 판정되고 무엇이 기록되는지

### 솔루션

- [AI 에이전트 운영](https://www.alpacax.com/kr/solutions/ai-agent-ops/): Claude, Codex, SRE 봇에 범위가 지정된 프로덕션 접근을 부여하고 모든 명령을 실행 전에 판정
- [팀 접근](https://www.alpacax.com/kr/solutions/team-access/): 수백 대 서버를 로그인 하나로, 설치 없이 브라우저에서 터미널, 파일, 에디터, 터널까지
- [CI/CD 및 자동화](https://www.alpacax.com/kr/solutions/automation/): 파이프라인에 자체 ID를 부여하고, 기본 거부 허용 목록과 IP·시간 제한, 로테이션 정책으로 통제
- [벤더 접근 관리](https://www.alpacax.com/kr/solutions/vendor-access-management/): 세션과 함께 만료되는 시간 제한·기록형 외부 인력 접근

### 개발자 도구

- [문서](https://docs.alpacax.com/): 설치 가이드, API 레퍼런스, MCP 연동, CLI 사용법
- [CLI 레퍼런스](https://docs.alpacax.com/reference/cli/): 동일한 정책 아래의 터미널 접근, 파일 전송, 일괄 명령
- [MCP 연동](https://www.alpacax.com/kr/alpacon/features/): Claude, Codex, VS Code를 위한 Model Context Protocol 인터페이스

### 통합

- 인증: SAML SSO, LDAP, MFA (생체 인증, 하드웨어 키)
- CI/CD: GitHub Actions, Docker 이미지, REST API 서비스 토큰
- 승인 및 알림: Slack, 이메일, 브라우저 푸시, 웹훅 (Discord, Microsoft Teams, Telegram)

### 콘텐츠

- [블로그](https://www.alpacax.com/kr/blog/): AI 에이전트 실행 제어에 관한 엔지니어링, 사고 분석, 제품 업데이트
- [뉴스룸](https://www.alpacax.com/kr/newsroom/): 회사 소식, 행사, 마일스톤
- [AI 에이전트로 php2shell을 6분 만에 패치한 방법](https://www.alpacax.com/kr/blog/how-we-patched-php2shell-in-six-minutes-with-an-ai-agent/)
- [승인 문서가 설명하는 모델은 이미 바뀌었을 수 있습니다](https://www.alpacax.com/kr/blog/the-model-you-approved-isnt-the-model-running-today/)
- [루프재킹: 승인한 동작과 실행된 동작의 불일치](https://www.alpacax.com/kr/blog/loopjacking-what-you-approved-isnt-what-ran/)
- [Product update: 요청이 들어온 경로와 알림 설정](https://www.alpacax.com/kr/blog/product-update-how-a-request-arrived-and-what-reaches-you/)
- [에이전틱 AI란 무엇인가: 세 가지 기준](https://www.alpacax.com/kr/blog/what-is-agentic-ai-the-definition-that-actually-holds-up/)
- [AI 에이전트가 벌인 PaperCut 대규모 침해](https://www.alpacax.com/kr/blog/inside-the-ai-agent-mass-exploitation-of-papercut-ng-mf/)
- [크라우드스트라이크 두 사고가 가리키는 커널 위험](https://www.alpacax.com/kr/blog/two-crowdstrike-failures-one-kernel-mode-security-risk/)
- [MCP 보안은 설치 검토로 끝나지 않습니다](https://www.alpacax.com/kr/blog/mcp-security-cant-stop-at-the-install-review/)
- [관리자 도구를 쥔 AI 에이전트, 접근 제어의 사각지대](https://www.alpacax.com/kr/blog/what-ai-agent-access-control-actually-has-to-check/)
- [Supabase와 Vercel에서 직접 운영으로, 실제로 달라지는 것들](https://www.alpacax.com/kr/blog/supabase-and-vercel-or-your-own-servers/)
- [제이드퍼퍼, 최초로 문서화된 AI 에이전트 랜섬웨어 공격](https://www.alpacax.com/kr/blog/inside-jadepuffer-the-first-documented-ai-agent-ransomware-attack/)
- [BNOW는 지금 팀으로 프로덕션 장애를 진단하고 해결합니다](https://www.alpacax.com/kr/blog/bnow-fixing-production-without-a-devops-team/)
- [사람이 막아 낸 AI 에이전트, 통제는 아니었습니다](https://www.alpacax.com/kr/blog/a-human-caught-the-ai-agent-thats-not-a-control-you-can-rely-on/)
- [AI 공격 프레임워크, 6시간도 안 돼 수천 건의 자격증명 수집](https://www.alpacax.com/kr/blog/ai-agent-framework-harvested-thousands-of-credentials-in-under-6-hours/)
- [은행의 모델 리스크 관리에서 제외된 AI 에이전트](https://www.alpacax.com/kr/blog/bank-regulators-wrote-ai-agents-out-of-model-risk-review/)
- [MCP 최소 권한은 두 단계에서 작동합니다](https://www.alpacax.com/kr/blog/mcp-least-privilege-has-two-layers-not-one/)
- [Product update: 다시 확인되는 것과 기록으로 남기는 것](https://www.alpacax.com/kr/blog/product-update-what-gets-re-checked-and-what-gets-written-down/)
- [API 토큰이 모든 명령을 실행할 수 있는 것은 아닙니다](https://www.alpacax.com/kr/blog/a-token-that-can-reach-your-api-isnt-cleared-to-run-any-command/)
- [상시 권한 없음, 내 환경에서 확인하는 7가지 질문](https://www.alpacax.com/kr/blog/zero-standing-privilege-a-7-question-audit-for-your-stack/)
- [MFA 피로는 사용자가 아니라 확인 범위의 문제입니다](https://www.alpacax.com/kr/blog/mfa-fatigue-is-a-scoping-problem-not-a-user-problem/)
- [AI 컴플라이언스 도구가 계약직 에이전트 리스크를 놓치는 이유](https://www.alpacax.com/kr/blog/ai-compliance-software-doesnt-solve-contractor-ai-agent-risk/)
- [Metabase 비밀번호 재설정이 관리자 권한으로 이어진 CVSS 10.0 취약점](https://www.alpacax.com/kr/blog/a-cvss-10-0-bug-turned-metabases-password-reset-into-full-admin/)
- [PAM 신원 매핑에서 마지막 단계가 조용히 실패하는 이유](https://www.alpacax.com/kr/blog/identity-mapping-in-pam-four-handoffs-one-fails-silently/)
- [AI 벤더 체크리스트: 채점할 실행 시점 질문 6가지](https://www.alpacax.com/kr/blog/ai-vendor-questionnaire-checklist-6-runtime-questions-to-score/)
- [접근 제어만으로는 AI 에이전트 거버넌스가 안 됩니다](https://www.alpacax.com/kr/blog/ai-agent-governance-means-more-than-access-control/)
- [검증 가이드는 나와도 규칙으로 만드는 건 사람의 몫입니다](https://www.alpacax.com/kr/blog/verification-guidance-exists-turning-it-into-a-rule-is-still-your-job/)
- [Product update: 명령이 실행되기 전에 정해두는 것들](https://www.alpacax.com/kr/blog/product-update-what-gets-decided-before-a-command-runs/)
- [AI 거버넌스 파일럿이 조직 규모가 커지면 무너지는 이유](https://www.alpacax.com/kr/blog/why-ai-governance-pilots-dont-survive-scale/)
- [별도 채널 승인: 에이전트 실행 채널은 승인 채널이 될 수 없습니다](https://www.alpacax.com/kr/blog/out-of-band-approval-why-an-agents-own-channel-cant-approve-itself/)
- [HIPAA 90일 후 감사: 세션 만료일만 봐서는 놓치는 것](https://www.alpacax.com/kr/blog/hipaa-day-90-audit-what-session-expiry-misses/)
- [쿠버네티스 접근 제어는 명령어를 판단하지 않습니다](https://www.alpacax.com/kr/blog/kubernetes-access-control-governs-access-not-commands/)
- [세션 단위 sudo: root 권한을 세션에 묶는 이유](https://www.alpacax.com/kr/blog/session-scoped-sudo-bind-os-level-root-to-a-session-not-a-sudoers-file/)
- [EU AI Act 고위험 항목 시한 연기, 바뀌지 않은 의무](https://www.alpacax.com/kr/blog/eu-ai-act-high-risk-deadline-moved-to-2027-what-didnt/)
- [보안 도구가 공격 경로가 된 2026년 CVE 2건](https://www.alpacax.com/kr/blog/two-2026-cves-turned-security-tools-into-the-attack-path/)
- [Product update: 감사 로그에서 비밀 값은 빼고, 접속 사각지대는 채웠습니다](https://www.alpacax.com/kr/blog/product-update-what-the-audit-log-stops-keeping-and-stops-missing/)
- [SSO와 MFA는 세션 동작까지 확인하지 않습니다](https://www.alpacax.com/kr/blog/session-layer-catches-what-sso-misses/)
- [탈취된 자격증명: 권한만 있고 목적은 없습니다](https://www.alpacax.com/kr/blog/credential-theft-entitlements-without-purpose/)
- [명령어 필터링이 뚫리는 두 가지 이유: GTFOBins, 그리고 매칭한 내용을 다시 쓰는 셸](https://www.alpacax.com/kr/blog/command-filtering-fails-two-ways-gtfobins-and-a-shell-that-rewrites-what-you-matched/)
- [오프보딩 접근 권한 회수: 퇴사한 엔지니어의 계정 밖에 남는 것](https://www.alpacax.com/kr/blog/offboarding-access-revocation-what-a-departing-engineer-leaves-running/)
- [AI 에이전트 실행 제어: 정의와 AI 네이티브 PAM에서 필요한 이유](https://www.alpacax.com/kr/blog/ai-agent-execution-control-what-it-is-and-why-ai-native-pam-needs-it/)
- [AgentForger: 링크 하나로 실제 직원의 인증 세션을 빌려 위조된 AI 에이전트](https://www.alpacax.com/kr/blog/agentforger-how-one-link-forged-a-rogue-ai-agent-with-a-borrowed-employee-session/)
- [2026년 AI 벤더 리스크 평가: 기존 질문지에 빠진 런타임 질문](https://www.alpacax.com/kr/blog/ai-vendor-risk-assessment-the-runtime-questions-your-questionnaire-is-missing/)
- [ISO 42001 Stage 1 체크리스트: 심사 전에 준비해야 할 문서](https://www.alpacax.com/kr/blog/iso-42001-stage-1-checklist-the-documentation-to-have-ready-before-your-audit/)
- [설계로 구현하는 상시 권한 없음: 토큰이 아니라 세션의 속성으로](https://www.alpacax.com/kr/blog/zero-standing-privilege-by-design-a-property-of-the-session-not-the-token/)
- [ISO 42001 심사 체크리스트: 정책 문서만으로는 충족할 수 없는 통제 항목](https://www.alpacax.com/kr/blog/iso-42001-audit-checklist-the-controls-a-policy-document-cant-satisfy/)
- [정렬만으로는 부족합니다: 오정렬된 에이전트의 행동을 제한하는 런타임 실행 제어](https://www.alpacax.com/kr/blog/alignment-isnt-enough-containing-a-misaligned-agent-takes-runtime-execution-control/)
- [문서 속 AI 거버넌스와 실행 중의 거버넌스](https://www.alpacax.com/kr/blog/ai-governance-on-paper-vs-governance-during-the-task/)
- [Product update: 승인에 판단 근거를 더했습니다](https://www.alpacax.com/kr/blog/product-update-the-judgment-behind-every-approval/)
- [자율 AI 에이전트가 Hugging Face를 침해했습니다: 킬체인과 실행 제어가 개입할 수 있는 지점](https://www.alpacax.com/kr/blog/an-autonomous-ai-agent-breached-hugging-face-heres-the-kill-chain/)
- [LegacyHive: 패치로 해결할 수 없는 Windows 제로데이](https://www.alpacax.com/kr/blog/legacyhive-a-windows-zero-day-you-cant-patch-your-way-out-of/)
- [AutoJack: 웹페이지 하나, MCP 소켓 하나로 호스트까지](https://www.alpacax.com/kr/blog/autojack-one-webpage-one-mcp-socket-host-level-access/)
- [커맨드 로그는 에이전트가 실제로 무엇을 했는지 말해주지 않습니다.](https://www.alpacax.com/kr/blog/a-command-log-wont-tell-you-what-your-agent-actually-did/)
- [JIT 접근 제어는 사람을 위한 것입니다. AI 에이전트에겐 무엇이 필요할까요?](https://www.alpacax.com/kr/blog/jit-access-control-is-for-humans-what-should-we-have-for-ai-agents/)
- [AI 에이전트를 사용한다고 보안을 포기해야 하는 것은 아닙니다](https://www.alpacax.com/kr/blog/adding-ai-agents-shouldnt-mean-adding-more-security-exposure/)
- [게이트는 작동했다. 그래도 털렸다.](https://www.alpacax.com/kr/blog/the-gate-worked-the-database-was-still-dumped/)
- [Work Sessions: CISO에게 정말 필요한 감사 단위](https://www.alpacax.com/kr/blog/work-sessions-the-audit-unit-your-ciso-needs/)

### 회사

- [회사 소개](https://www.alpacax.com/kr/about/): 팀과 미션; 2022년 설립; 멘로파크·서울·대전 오피스; Rabbit Ventures와 KAIST Ventures가 투자했고, POSCO IMP는 파트너 프로그램
- [채용](https://www.alpacax.com/kr/careers/): AlpacaX 채용 공고

### 신뢰 및 보안

- [트러스트 센터](https://trust.alpacax.com/): SOC 2 Type 2 (진행 중), 실시간 통제 현황, 리포트, 서브프로세서
- [보안 개요](https://www.alpacax.com/kr/security/): 아키텍처, 판정, 취약점 공개 정책, 가용성

### 기타

- [GitHub](https://github.com/alpacax): 오픈소스 alpamon, alpacon-cli, alpacon-mcp
- [앱](https://alpacon.io/): 기존 Alpacon 워크스페이스 로그인