Skip to main content
AI Inventory shows where ZeroPath detects AI SDKs, model files, MCP server packages and supported configurations, and other agent tooling in your repositories. Use it to decide which components need review, inspect the package or file behind a detection, and share the inventory with your team.

Review AI components

In Assets, open Repositories, select a repository, and open Inventory → AI components.
  1. Select the branch in Inventory. Expand the repository row if needed to see its components.
  2. Search by package, provider, or path. Use Kind, Usage, Model type, Provider, or Detection source to narrow the results.
  3. Check the scan date, manifest, or file path to see where each component was detected and decide what to review.
The view shows detections from the selected branch’s most recent successful scan with AI components. An older inventory can stay visible after a scan with no detections. Check the scan date when reviewing recent changes.

What gets inventoried

  • AI and ML dependencies include LLM SDKs, agent and ML frameworks, ML runtimes, vector databases, guardrails, observability libraries, speech and audio libraries, and tokenizers.
  • Models and data include detected model files such as GGUF, ONNX, and SafeTensors, local model configurations, and datasets.
  • Agent tooling includes MCP server packages and supported configurations, agent configurations, instruction files, skills, and prompt templates.
ZeroPath detects these components from recognized dependencies and repository files. Usage classifies detections as Runtime, Dev Tooling, or Ambiguous based on available repository evidence. Component details can include provider, version, and detection source when available. Dependency-backed components can also include direct or transitive status and reachability evidence from the linked package.

Export the AI-BOM

In the repository’s Inventory → AI components view, choose Export ML-BOM, then Generate SBOM. The export includes the AI components from the selected repository scan in CycloneDX 1.6 JSON. Use this file to share the scan’s inventory for review or reporting. If the selected inventory comes from a standalone SCA scan, the export needs a completed full scan on the same branch. Run a full scan on that branch and try again if ZeroPath asks for one. See SBOM exports for other formats and export requirements.

Enable AI Inventory

Under Full scans → Full Scan Issue Types, enable SCA and AI Inventory. Both modules are enabled by default for full scans. Turn off Custom Rules Only so the full scan can collect AI components during SCA. Standalone SCA scans also collect AI components when the AI Inventory module is enabled. If you see no components, clear the filters, check the branch, and run a new full scan with both modules enabled. See scanner settings for how to change the modules.

API access

  • POST /api/v1/sca/ai-inventory/repositories/search lists repositories with detected components and counts by kind. Use branch to scope the results. A single-repository search can also return exportSelection for ML-BOM export.
  • POST /api/v1/sca/ai-inventory/components/search lists component rows for a repository and scan.
  • POST /api/v1/sca/ai-inventory/providers/search lists detected providers for the repository and branch scope.