Securing the AI Coder: JFrog and Anthropic Tackle Supply Chain Risk

SUNNYVALE, CA – June 10, 2026 – In a move that directly confronts the burgeoning security risks of AI-driven software development, JFrog has partnered with Anthropic to embed its software supply chain platform directly into the Claude Code AI assistant. The new plugin, available immediately, represents a critical step toward taming the Wild West of autonomous code generation, providing enterprises with a much-needed layer of governance and visibility over their increasingly automated development pipelines. As AI agents evolve from helpful assistants to active participants in creating and deploying software, this collaboration signals a market-wide shift from reactive security scans to proactive, real-time governance.

Taming the Autonomous Coder

The rise of AI coding agents has been a phenomenal boon for developer productivity, but it has introduced a significant and often overlooked vulnerability. These agents, in their quest to build software at speed, pull in dependencies, libraries, and code snippets from a vast digital commons, frequently without any context for their security posture or licensing compliance. This creates a massive blind spot for organizations.

"AI agents are active participants in the software supply chain, making decisions about dependencies, builds, and deployments – but most of them are doing it blind, without any supply chain context," explained Yoav Landman, Co-Founder and CTO of JFrog. "This is often how malicious packages, vulnerabilities, and ungoverned AI assets enter production today, exposing organizations to software supply chain attacks."

The scale of the problem is staggering. JFrog reports that its platform now manages over 18 billion software artifacts, a 136% increase from the previous year, with much of this surge attributed to the prolific output of AI coding tools. Each artifact represents a potential entry point for risk. The concern is echoed by AI developers themselves. Anthropic, the creator of Claude, has publicly highlighted the need for collective investment in agent-specific security, noting that "as agents grow more capable, attack surfaces are constantly shifting." This partnership is a direct answer to that call, aiming to instill guardrails before a major, AI-assisted supply chain breach makes headlines.

A System of Record for AI-Generated Code

The JFrog Platform plugin for Claude Code is designed to move security from an afterthought to an integral part of the AI-assisted coding process. Rather than running scans after code has been written and integrated, the plugin provides real-time governance directly within the developer's workflow. Using natural language prompts, a developer can now ask their Claude agent to not only write code but also to ensure its components are secure and compliant from the outset.

Functionally, the integration allows the AI agent to tap into the JFrog Platform's deep repository of security intelligence. When an agent suggests a software package, the plugin can instantly scan it for known vulnerabilities (CVEs), check its license against company policy, and verify its provenance. This "shift-left" approach, embedded directly into the AI's decision-making process, prevents risky components from ever entering the codebase.

The plugin provides four key benefits:
* Real-time, Upstream Governance: Security and compliance checks happen as code is written, eliminating the risky and time-consuming manual handoffs that traditionally slow down release cycles.
* Accelerated DevOps Workflows: Repetitive platform management tasks, such as repository creation or project provisioning, can be delegated to the agent through JFrog Platform Skills, freeing up engineers to focus on innovation.
* Governed Access: The integration ensures that agents and developers only pull from verified, secure sources, blocking rogue access to sensitive internal data or the use of unapproved components.
* Strengthened Auditability: It creates an end-to-end, traceable log from the initial source commit to the final build artifact. When an audit is required or an incident occurs, security teams can trace the lineage of every component in minutes, not days.

This functionality addresses a tangible need felt within the developer community. The recent emergence of community-built tools designed to intercept and block compromised packages in AI workflows demonstrates a clear, grassroots demand for exactly the kind of enterprise-grade solution this partnership formalizes.

The Rise of Agentic DevSecOps

While the Claude Code integration is significant, it represents just one piece of JFrog's much broader strategic vision: to become the universal system of record for governance in a multi-agent world. The company is betting that development teams won't standardize on a single AI agent but will instead use a mix of tools, including Claude, GitHub Copilot, Cursor, and others. In this fragmented landscape, consistent governance becomes paramount.

To achieve this "agent universality," JFrog is building out a three-layered architecture:
1. JFrog Platform Skills: These give any connected agent deep, domain-specific knowledge of the JFrog Platform, enabling complex operations like vulnerability scanning and provenance checks through simple natural language commands.
2. JFrog MCP (Model Context Protocol) Tools: These provide standardized access to security, compliance, and artifact data, ensuring consistent governance regardless of which agent initiates a request.
3. Agent-Native Plugins: Starting with Claude Code and extending to other environments like VS Code Copilot, these plugins bring the full power of the JFrog Platform into each agent’s native environment with simple authentication.

This strategy positions the platform not as another tool, but as a foundational governance fabric that travels with the developer across their entire toolchain. It's a play to own the crucial trust layer for what the industry is beginning to call "Agentic DevSecOps," where autonomous agents are managed and secured by a central policy engine.

Shifting the Paradigm for Secure Development

By embedding security context directly into the AI agent, the JFrog-Anthropic collaboration challenges the traditional model of software security. For decades, the paradigm has been to build first and scan later. This approach is untenable in an era where AI can generate and integrate code faster than any human security team can review it.

The integration offers a new model where security and compliance are not gates, but guardrails. "AI-enabled innovation cannot come at the expense of security or compliance," Landman stated. "Enterprises need a universal system of record with real-time control and visibility into the decisions these agents make, that's what this integration enables."

This move places JFrog in a competitive position within the burgeoning AI security landscape, differentiating its artifact-centric approach from other DevSecOps players who may focus on different stages of the lifecycle. By making the binary artifact the central point of control, the company aims to provide definitive, auditable proof of compliance for every component, whether written by a human or an AI. For enterprises navigating the dual pressures of rapid innovation and stringent regulatory requirements, this promise of balancing speed with safety is a powerful proposition. The ultimate goal is to create a software supply chain where AI is not a source of unmanaged risk, but a secure and compliant accelerator of progress.