New Agent Governance & Package Firewall capabilities close critical visibility and governance gaps in AI coding agents and software supply chains
Endor Labs, the leader in agentic application security, today announced two new capabilities as part of AURI. The first, Agent Governance, gives security and engineering teams real-time visibility and policy enforcement over AI coding agents – what they integrate and run, what they do, and what they produce. The second, Package Firewall, blocks malicious packages before they can reach developer workstations or agent-driven workflows. Together, these capabilities give organizations the controls to treat the agentic development environment as the production infrastructure it has now become.
AI is accelerating developers and attackers alike. Developers are shipping faster than ever with coding agents that write code, execute commands, install packages, and interact with critical systems autonomously, often at a pace and scale that legacy application security tools were never designed to handle.
Attackers are moving just as fast. Endor Labs’ reports about malware in open source ecosystems found a 14x increase in malware advisories over the last two years, and that 92% of all npm maintainer account takeovers — where an attacker compromises a legitimate account to distribute malware — ever recorded happened in 2025 alone. Increasingly, these attacks leverage open source packages to compromise AI coding agents, from the recent compromise of the Bitwarden CLI to the Nx build platform compromise, where attackers weaponized Claude and Gemini CLIs to exfiltrate developer secrets.
The pattern across these incidents is consistent: supply chain attacks are no longer confined to application code. They now extend into the infrastructure layer — the agents, workstations, and CI systems that generate and deliver software. Security teams have to defend not only what gets written, but the integrity of the systems writing it.
“AI coding agents now operate with the same access as your most trusted engineers, and most organizations have no idea what those agents are actually doing,” said Varun Badhwar, CEO & Founder, Endor Labs. “CISOs need improved visibility, alerts, and a control plane for the AI agents operating within their environments. With Agent Governance and Package Firewall, we’re giving them the ability to see and control what AI agents are actually doing, without slowing engineering velocity.”
The result is a dangerously expanding blind spot for security and engineering teams. Nearly 70% of CISOs report limited visibility into how AI is being used across their environments, and engineering leaders often can’t see which agents and models their teams are running. That makes it difficult to assess exposure, enforce policy, or respond to incidents.
“Enterprise engineering and security leaders are not asking whether to adopt agentic coding; they are asking how to do it at scale with the trust the business requires,” said Brian McCarthy, President, Global Revenue and Field Operations, Cursor. “We have invested heavily in tools, security controls, and governance, along with partnerships, including with Endor Labs, that let security teams see what every agent is doing, enforce policies across workstations, and ship with confidence. The result is developers moving at full speed with the guardrails enterprises need.”
“By leveraging integrations such as Model Context Protocol (MCP) servers, Endor Labs brings its security intelligence directly into the Gemini ecosystem — providing the real-time guardrails and attribution that allow Google Cloud customers to scale AI-native workflows securely,” said Vikas Anand, Director, Product Management, Google Cloud.
Agent Governance gives organizations full visibility into AI activity across developer workstations and cloud environments, creating a system of record for how agents operate. Agent Governance natively integrates directly into the hook systems that Cursor, Claude Code, Gemini CLI, and other coding agents expose — without requiring separate infrastructure, per-developer configuration, or changes to how developers work.
With this capability, security teams can:
- Inventory AI agents, models, and tools across environments
- Track MCP server usage and agent “skills” to understand behavior and risk
- Detect and block dangerous actions in real time across prompts, commands, file access, and tool usage
Package Firewall addresses one of the most immediate risks in AI-driven development: the uncontrolled ingestion of external dependencies. The capability intercepts every package installation request – from a developer or an AI agent – and blocks malicious or vulnerable packages before they reach the environment. This is all accomplished without disrupting developer workflows or CI/CD pipelines. Key features include:
- Real-time scanning of every open-source package across npm, PyPI, NuGet, Maven, and other major ecosystems
- Policy enforcement for vulnerabilities, licensing, and package trust
- Cooldown periods to reduce exposure to newly published malicious packages
- Seamless deployment via artifact registries or proxy, without developer disruption
AURI by Endor Labs, first introduced this past March, is the security harness for agentic development. It is powered by the patented code context graph that builds a searchable model of your architecture, services, code, open-source dependencies, and container images. AURI uses this context to understand code behavior, tailoring security findings to the context of your application. It is fully integrated with AI coding agents via Hooks, Skills, MCP, or CLI, so risks can be identified and resolved before they propagate into production.
Agent Governance is available in private preview. Package Firewall is available today as part of the Endor Labs AURI platform. To learn more, please visit: https://www.endorlabs.com/platform.
Explore AITechPark for the latest Artificial Intelligence News advancements in AI, IOT, Cybersecurity, AITech News, and insightful updates from industry experts!