JFrog AI Catalog

JFrog AI Catalog is the AI governance layer of the JFrog Platform that registers, scans and enforces policy on AI models, MCP servers, agent skills and plugins before coding agents or applications use them. It is built for platform engineering and AppSec teams securing agentic development on Artifactory.

JFrog AI Catalog is the AI governance and security layer of the JFrog Platform, from the Nasdaq-listed software supply chain company founded in 2008 that builds Artifactory. It launched at swampUP 2025 in September 2025 as a secure model registry and has since grown into what JFrog calls an AI control plane that treats models, MCP servers, agent skills, coding-agent plugins and agent packages as first-class artifacts alongside ordinary binaries. The catalog adds a Model Registry, MCP Registry, Agent Skills Registry, Agent Plugins Registry and an Agent Package Manager registry built on the Microsoft-led APM standard, with policy and security scanning applied through JFrog Xray and curation gates that block risky assets before they enter the supply chain. At swampUP 2026, announced September 2-3, 2026, JFrog added Agent Guard, which routes coding agents such as Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Kiro, Devin and VS Code through an authenticated bridge that enforces project-scoped allow and deny rules; AI Asset Scanning, a semantic scan of markdown, skill scripts and instruction sets for malicious intent; Shadow AI Detection for unapproved models already in repositories; and Traffic Controller, backed by Cloudflare, Netskope and Zscaler, which blocks direct calls to public registries. The same release added Zero-Touch Remediation, AppTrust governance features and a Wiz integration. In June 2026 JFrog and the NanoClaw agent project also shipped an integration that forces agent package installs through vetted JFrog registries. AI Catalog and Agent Guard are included in the Enterprise+ tier and sold as add-ons on lower tiers, on SaaS across AWS, Azure and Google Cloud or self-managed. Independent reviews of the underlying Artifactory platform praise its breadth but flag cost, complexity and a steep learning curve.

Category

Governance & Security

Pricing

Subscription, Usage-based

Target Market

CTOs, CISOs, Platform Engineering Leaders, Application Security Teams, DevOps Teams

Deployment

Hybrid, Multi-cloud, Self-hosted

Founded

2008

Customers

Thousands of JFrog Platform customers (vendor-reported); AI Catalog adoption not disclosed

• ✓
  AI asset registries

  Stores models, MCP servers, agent skills, coding-agent plugins and agent packages as versioned artifacts with policy applied on every pull.

• ✓
  Agent Guard

  Routes coding agents like Claude Code, Cursor and Copilot through an authenticated bridge enforcing project-scoped allow and deny policies.

• ✓
  AI Asset Scanning

  Semantically scans markdown, skill scripts and instruction sets to catch malicious behaviour hidden in AI assets before agents run them.

• ✓
  Shadow AI Detection

  Finds unapproved AI models already present in repositories and checks them for vulnerabilities and malicious code.

• ✓
  Curation gates

  Blocks risky or non-compliant models and packages before they enter the organization's software supply chain, rather than after deployment.

• ✓
  Traffic Controller

  Blocks direct calls to public registries at the network layer and reroutes traffic through Artifactory, supported by Cloudflare, Netskope and Zscaler.

• ✓
  Agent Package Manager support

  Integrates Microsoft's APM standard so agent dependencies are packaged, version-pinned and resolved only from approved sources.

• •
  Securing coding-agent rollouts

  Platform teams let thousands of developers use Claude Code or Cursor while Agent Guard restricts plugins and MCP servers to vetted ones.

• •
  Stopping malicious package installs

  Agents that try to install a compromised library receive a policy error and are steered to an approved safe version instead.

• •
  Shadow AI inventory

  Security teams scan existing repositories to surface unapproved models and agent assets, then decide what to approve, remediate or remove.

• •
  Approved model catalog

  Organizations publish a governed list of external API models, Hugging Face models and internal models that each project may consume.

• •
  Compliance evidence for AI-built software

  AppTrust traceability records which prompts, assets and approvals led to a release, supporting audits under CRA and NIST SSDF.

JFrog publishes list prices for the platform but not for the AI Catalog or Agent Guard add-ons. Both are bundled only in the custom-priced Enterprise+ tier; on SaaS Pro ($150/month) AI Catalog is an add-on and Agent Guard is unavailable, and on Enterprise X (from $950/month SaaS or $51,000/year self-managed) both are add-ons. SaaS plans are metered on consumption beyond included storage; competitor CloudRepo notes data transfer, including CI downloads, counts toward that consumption.

This page was written from 10 sources, 6 on domains other than jfrog.com.

1. 1.jfrog.com — ai catalogvendor
2. 2.jfrog.com — the evolution of jfrog ai catalogvendor
3. 3.jfrog.com — jfrog embeds security into the agentic workforcevendor
4. 4.jfrog.com — pricingvendor
5. 5.docs.jfrog.com — jfrog ai catalog overview
6. 6.devops.com — jfrog moves to secure agentic engineering workflows
7. 7.securitybrief.co.uk — jfrog launches ai era security tools for software supply
8. 8.venturebeat.com — nanoclaw and jfrog launch immune system to block ai agents f
9. 9.peerspot.com — jfrog artifactory pros and cons
10. 10.cloudrepo.io — jfrog artifactory hidden costs