# Knostic

> Knostic is the control plane for the agentic AI code stack — secure Claude Code, Cowork, MCP servers, and skills without slowing developers down.

Knostic builds security for the agentic AI code stack: Kirin (runtime guardrails for AI coding agents), AgentMesh (supply-chain scanning for agent skills, MCP servers and extensions) and OpenAnt (open-source LLM-powered vulnerability discovery). Blog posts and resources are the best sources for how the products work and the research behind them.

## Products

- [Kirin](https://www.knostic.ai/kirin): Secures AI coding assistants from prompt injection, rogue behavior and unsafe execution
- [AgentMesh](https://www.knostic.ai/agentmesh): Discovers, tracks and scans AI agent skills, MCP servers and VS Code extensions for prompt-injection and supply-chain threats
- [OpenAnt](https://www.knostic.ai/openant): Open-source, LLM-powered vulnerability discovery that finds exploitable bugs and eliminates false positives
- [Pricing](https://www.knostic.ai/pricing): Free plans for individual developers, per-user team plans, custom enterprise plans

## Solutions

- [Shadow AI Discovery & Governance](https://www.knostic.ai/solutions/shadow-ai-discovery-governance): Shadow AI discovery and governance: find every AI agent, extension, MCP server, and skill in use on your endpoints and bring it under one policy.
- [Prompt Injection Defense](https://www.knostic.ai/solutions/prompt-injection-defense): Prompt injection defense for AI agents: Kirin scans prompts and context and blocks the dangerous action an injection would cause, before it runs.
- [MCP Security Posture Management](https://www.knostic.ai/solutions/mcp-security-posture-management): MCP security posture management: inventory and validate every MCP server, enforce secure configuration, flag drift, and block rogue connectors.
- [AI Agent Security](https://www.knostic.ai/solutions/ai-agent-security): AI agent security: Kirin applies least privilege to agent actions, monitors every tool call, and blocks unsafe behavior before it executes.

## Use cases

- [Coding Assistants, Agents & MCP](https://www.knostic.ai/use-cases/coding-assistants-agents-mcp): Knostic secures AI coding assistants, autonomous agents, and MCP servers. Kirin enforces guardrails inside the agent; AgentMesh vets what it installs.
- [Employees & End Users](https://www.knostic.ai/use-cases/employees-end-users): Knostic secures Cowork agents, desktop assistants, and the skills employees install: shadow AI discovery, vetted plugins, and guardrails on risky actions.
- [Enterprise Control Plane](https://www.knostic.ai/use-cases/enterprise): Knostic's enterprise control plane sets agent security policy once and enforces it across every AI tool, with coverage dashboards, SIEM alerts, and audit.
- [Security Teams](https://www.knostic.ai/use-cases/security-teams): Knostic for security teams: block dangerous agent actions before they run, vet skills and MCP servers, discover shadow AI, and alert your SOC.
- [Governance & Compliance](https://www.knostic.ai/use-cases/governance-compliance): Knostic for governance and compliance: enforce AI acceptable-use policy inside every agent, detect drift, and keep audit-ready logs of every action.
- [IT & Collaboration](https://www.knostic.ai/use-cases/it-collaboration): Knostic for IT: discover every AI agent and plugin in use, vet skills and MCP servers before install, and manage agent policy from one console.
- [Product & Engineering](https://www.knostic.ai/use-cases/product-engineering): Knostic for engineering teams: enforce secure coding in AI assistants, block hallucinated packages and bad MCP servers, find real vulnerabilities with OpenAnt.
- [Data & Analytics](https://www.knostic.ai/use-cases/data-analytics): Knostic for data and analytics teams: block destructive agent queries, catch credentials and PII in prompts, and see every AI tool touching your data.
- [Customer Success & Support](https://www.knostic.ai/use-cases/customer-support): Knostic for customer support: govern what AI agents can read and do in CRMs and ticket systems, vet plugins, and catch PII before it leaves.
- [Red Teams & Pen Testers](https://www.knostic.ai/use-cases/red-teams-pen-testers): Knostic for red teams: test prompt injection, poisoned skills, and rogue MCP servers against real agents, and find exploitable bugs with OpenAnt.
- [Identity & Access Management](https://www.knostic.ai/use-cases/identity-access-management): Knostic for IAM: extend least privilege to AI agents, control MCP connections, catch credentials, and log every agent action with its identity.
- [CISO](https://www.knostic.ai/use-cases/ciso): Knostic for CISOs: shadow AI discovery, inline prevention of dangerous agent actions, one policy across every tool, and board-ready evidence.
- [CIO](https://www.knostic.ai/use-cases/cio): Knostic for CIOs: roll out AI agents company-wide in days, enforce one policy everywhere, and track adoption, risk, and coverage from one dashboard.
- [Executives & Board](https://www.knostic.ai/use-cases/executives-board): Knostic for executives and boards: see every AI agent in the company, stop dangerous actions before they run, and report prevented incidents with evidence.
- [Energy](https://www.knostic.ai/use-cases/energy): Knostic for energy: govern AI agents across critical infrastructure teams, block dangerous actions, vet components, and keep NERC CIP-ready logs.
- [Financial Services](https://www.knostic.ai/use-cases/financial-services): Knostic for financial services: block dangerous agent actions in production, enforce secure AI coding, and keep SOX, PCI, and DORA evidence.
- [Healthcare](https://www.knostic.ai/use-cases/healthcare): Knostic for healthcare: catch PHI before it leaves an AI agent, block destructive actions on patient systems, vet plugins, and keep HIPAA audit evidence.
- [Pharmaceuticals](https://www.knostic.ai/use-cases/pharma): Knostic for pharma: vet packages and skills in research pipelines, protect trial data and IP from leaving agents, and keep GxP-ready decision logs.
- [Government](https://www.knostic.ai/use-cases/government): Knostic for government: discover shadow and contractor AI, block dangerous agent actions, vet components, and keep NIST-aligned audit logs.

## Company

- [About](https://www.knostic.ai/about): Team, investors and research behind Knostic
- [FAQ](https://www.knostic.ai/faq): Common questions about deploying Knostic and its impact on developers
- [Glossary of AI terms](https://www.knostic.ai/glossary)
- [Events](https://www.knostic.ai/events): Dinners, workshops, webinars and conference sessions
- [Book a demo](https://www.knostic.ai/demo)
- [Contact](https://www.knostic.ai/contact)

## Resources

- [Resource Center](https://www.knostic.ai/resource-center): Solution briefs, guides and reports
- [Solution Brief: LLM Data Leakage](https://www.knostic.ai/resources/solution-brief-request): Get the Solution Brief & learn how Knostic helps enterprises detect and respond to LLM data leakage in AI Chatbots. Manage data exposure and your security.
- [Data Governance in the Age of LLMs](https://www.knostic.ai/resources/llm-data-governance-white-paper): Learn how Knostic prevents LLM-driven data leaks. Download the white paper to master AI-native governance and close knowledge-exposure gaps fast.
- [Copilot Oversharing Risks: Knostic and Purview](https://www.knostic.ai/resources/copilot-data-oversharing-solution-brief): Stop Microsoft Copilot data oversharing. See how Knostic enhances Purview with rapid risk visibility & AI-specific controls for secure Copilot adoption.
- [Knostic in AI TRiSM](https://www.knostic.ai/resources/ai-trism-solution-brief): Learn how Knostic enhances AI TRiSM. Secure enterprise LLMs like Copilot by defining & enforcing 'need-to-know' policies for knowledge-based access control.
- [LLM Jailbreak Prompts by Industry: A Hands-On Playbook](https://www.knostic.ai/resources/llm-jailbreak-prompts-playbook): Download Knostic’s industry-specific LLM jailbreak prompt Playbook. Test and stop AI oversharing in finance, healthcare, tech, retail, manufacturing & defense.
- [Safely Unlock Hidden Copilot Knowledge in Microsoft 365](https://www.knostic.ai/resources/unlock-copilot-knowledge): Unlock hidden knowledge with Copilot safely. Download Knostic’s white paper to see how adaptive security outperforms Microsoft Blueprint for rapid, secure rollouts
- [Unlock Copilot on E3 with Confidence](https://www.knostic.ai/resources/unlocking-copilot-on-e3): Learn how Knostic prevents LLM-driven data leaks. Download the white paper to master AI-native governance and close knowledge-exposure gaps fast.
- [Cyber Defense Matrix book](https://www.knostic.ai/resources/cyber-defense-matrix-book): Explore Knostic's resources on cybersecurity, AI risks, and access control solutions to enhance your enterprise's defense against evolving threats. Download our essential ebook today.
- [Copilot Readiness Assessment](https://www.knostic.ai/resources/copilot-readiness-assessment): Is your enterprise ready yet to deploy Copilot? Get started with a Microsoft 365 Copilot Readiness Assessment

## Blog

- [Blog index](https://www.knostic.ai/blog): Research, product news and practical guidance on securing agentic AI
- [An MCP that Reads Your Reddits and Runs Python Hidden in an Image](https://www.knostic.ai/blog/mcp-that-reads-your-reddits-and-runs-hidden-python-in-an-image): A PyPI MCP server for reading Reddit downloads a PNG, extracts Python hidden in its pixels and runs it on every tool call. How the chain works.
- [Auto-patch for OpenAnt: Generating trust in autonomous patching](https://www.knostic.ai/blog/auto-patch-for-openant)
- [The Duck Song: Malicious Solidity Extensions Hide a Cross-Platform Payload in an MP4](https://www.knostic.ai/blog/the-duck-song-malicious-solidity-extensions-hide-a-cross-platform-payload-in-an-mp4): Five malicious Solidity/Hardhat extensions; four hide a cross-platform payload in an MP4 that tries to install ScreenConnect or a persistent Go binary.
- [When “Auto-Signing” Sends Your Wallet: A Malicious MCP Server on npm](https://www.knostic.ai/blog/when-auto-signing-sends-your-wallet-private-key-to-a-remote-server-a-malicious-mcp-package-on-npm): A malicious npm package exposes wallet private keys, misleading users with false auto-signing claims, risking asset security. Learn how this vulnerability works.
- [SaassyCode Repackaged: Four New VS Code Extensions and a New Delivery Chain](https://www.knostic.ai/blog/saassycode-repackaged-four-new-vs-code-extensions-and-a-new-delivery-chain): Knostic reveals four new VS Code extensions that exploit SaassyCode's malicious code, showcasing evolving delivery methods and potential security risks.
- [CodeRelay: 12 VS Code Extensions Disguised as Developer Tools](https://www.knostic.ai/blog/coderelay-12-vs-code-extensions-disguised-as-developer-tools): CodeRelay: a coordinated campaign of 13 malicious VSIX packages across 12 VS Code extensions disguised as developer tools. Analysis, IOCs, and defenses.
- [Introducing AgentMesh Access Tiers: Building Together, Growing Together](https://www.knostic.ai/blog/introducing-agentmesh-access-tiers): Discover AgentMesh's new access tiers, offering enhanced features for developers and researchers while maintaining a robust free option for all users.
- [Lessons Learned from the Hugging Face Security Team](https://www.knostic.ai/blog/lessons-learned-from-the-hugging-face-security-team): Review key lessons from the Hugging Face Security Team on defending against autonomous AI adversaries and enhancing your security programs.
- [Analyzing Two Malicious VS Code Extensions Hidden Behind a WordPress Tool](https://www.knostic.ai/blog/analyzing-two-malicious-vs-code-extensions-hidden-behind-a-wordpress-tool): A technical investigation into two malicious VS Code extensions marketed as WordPress development tools, which utilize a recurring download-and-launch routine to execute external payloads on startup.
- [The Source Looked Clean. The Binary Wasn't.](https://www.knostic.ai/blog/the-source-looked-clean-the-binary-wasnt): A VS Code extension appeared safe but contained hidden malicious code, illustrating the risks of relying solely on source code for security.
- [OpenAnt: Read the LLM Vulnerability Paper](https://www.knostic.ai/blog/openant-read-the-llm-vulnerability-paper): Knostic's OpenAnt, al eading open-source tool that combines static analysis and LLM reasoning is now a paper.
- [Revoking Your Token Won't Save You: The VS Code Attack That Installs a Permanent GitHub Backdoor](https://www.knostic.ai/blog/revoking-your-token-wont-save-you-the-vs-code-attack-that-installs-a-permanent-github-backdoor): Malicious VS Code extensions can create a GitHub backdoor that survives token revocation and password changes, leaving developers vulnerable.
- [Knostic Featured Across Three Categories in CSA's Agentic AI Security Innovator Market Map](https://www.knostic.ai/blog/knostic-featured-across-three-categories-in-csas-agentic-ai-security-innovator-market-map): Knostic has been recognized by the Cloud Security Alliance in three key categories of its Agentic AI Security Innovator Market Map, showcasing its comprehensive approach to AI security.
- [SaassyCode Post-Disclosure Wave: Five New Extensions, 32,000+ Total Installs](https://www.knostic.ai/blog/saassycode-post-disclosure-wave-five-new-extensions-32000-total-installs): New findings reveal a surge in malicious VS Code extensions, with over 32,000 installs. One extension remains live, posing serious risks to developers.
- [Knostic’s AgentMesh Joins VirusTotal's Crowdsourced AI Program](https://www.knostic.ai/blog/knostics-agentmesh-joins-virustotals-crowdsourced-ai-program): Knostic's AgentMesh enhances VirusTotal's security with advanced AI threat intelligence, addressing the growing risks in the AI agent supply chain.
- [Update and Infect: How the SaassyCode Campaign Grew from Two Extensions to Nineteen](https://www.knostic.ai/blog/update-and-infect-how-the-saassycode-campaign-grew-from-two-extensions-to-nineteen): How one VS Code extension published clean, built an install base, then silently delivered malware via a routine update.
- [New Malicious VS Code Extension Backdoor: Remote Text Fetcher](https://www.knostic.ai/blog/new-malicious-vs-code-extension-backdoor-remote-text-fetcher): A VS Code extension disguised as a text utility executes attacker-controlled shell commands on every startup, caught by AgentMesh before reaching any users
- [New VS Code extensions attack campaign: SaassyCode - ManageRBLX & TrelloBlox](https://www.knostic.ai/blog/new-vs-code-extensions-attack-campaign-saassycode-managerblx-trelloblox): A new attack campaign targets VS Code users with malicious extensions, ManageRBLX and TrelloBlox, exposing serious security risks. Learn more about the threats.
- [Biggest Takedown of 2026 to Date? Glassworm: Targeting Developers at Scale](https://www.knostic.ai/blog/glassworm-targeting-developers-at-scale): CrowdStrike took down Glassworm, the botnet built to compromise software developers. Here's why the defense has to live in the IDE, and how Kirin does it.
- [GitHub Supply Chain Breach: Nx Console 18.95.0 (Analysis)](https://www.knostic.ai/blog/github-supply-chain-breach-nx-console-18.95.0-analysis): Nx Console v18.95.0 on OpenVSX auto-executed code from a pinned nrwl/nx commit on activation. Source-verified analysis of the Shai-Hulud VSIX.
- [Agentic Threat Intelligence Feed - VS Code Extensions](https://www.knostic.ai/blog/agentic-threat-intelligence-feed-vs-code-extensions): Knostic's free agentic threat intel: 4 source-validated VS Code extension findings (May 13–20, 2026), including BCAI Rosetta with 13,480 installs on Cursor.
- [The GitHub Breach and VS Code Extensions: Threat Intelligence and Agentic Defense](https://www.knostic.ai/blog/github-teampcp-breach-developer-supply-chain): The GitHub TeamPCP incident wasn't a one-off. It's the predictable result of an unmonitored developer agent surface. Here's what CISOs should do next.
- [The "AI Vulnerability Storm": A CISO's Playbook for the Mythos Era](https://www.knostic.ai/blog/the-ai-vulnerability-storm-a-cisos-playbook-for-the-mythos-era): OpenAnt: Knostic's open-source, LLM-based vulnerability discovery tool for proactively identifying security flaws while minimizing false positives and negatives. Download for free from GitHub.
- [Free LLM-based Vulnerability Scans for Open Source Projects](https://www.knostic.ai/blog/oss-scan): Discover how OpenAnt, a free LLM-based tool, transforms vulnerability scanning for open source projects, leveling the playing field for defenders against AI-powered attacks.
- [OpenAnt: Open Sourcing Knostic’s LLM-based Vulnerability Discovery Product](https://www.knostic.ai/blog/openant): OpenAnt: Knostic's open-source, LLM-based vulnerability discovery tool for proactively identifying security flaws while minimizing false positives and negatives. Download for free from GitHub.

## Optional

- [Sitemap](https://www.knostic.ai/sitemap-index.xml)
- [Data Privacy Agreement](https://www.knostic.ai/data-privacy-agreement)
- [Website Privacy Policy](https://www.knostic.ai/privacy-policy)
- [Terms of Service](https://www.knostic.ai/terms-of-service)