Inventory
The Inventory page gives you a complete, filterable list of every AI asset discovered across your organization's endpoints, with risk profiles and governance coverage insights.
What is the Inventory?
The Inventory page gives you a complete, filterable list of every AI asset discovered across your organization's endpoints. Use it to review what software exists in your environment, understand its risk profile, and identify gaps in your governance coverage.
Reading the Inventory Table
Each row in the inventory represents a single unique asset. The table columns are:
| Column | Description |
|---|---|
| Name | The name of the discovered asset |
| Asset Type | The category of the asset — Application, AI Skill, Browser Extension, IDE Plugin, MCP Server, and more |
| Status | The governance status of the asset — Ungoverned for assets not yet reviewed by your team |
| Risk Level | The assigned risk rating: Critical, High, Medium, or Low |
| Endpoints | The number of devices on which this asset has been discovered |
Asset Types
Surface classifies every discovered asset into one of the following types:
Application Desktop or web applications with AI capabilities. Examples include productivity tools, AI writing assistants, and AI-enhanced communication platforms.
AI Skill AI-powered scripting frameworks, agent toolkits, or reusable AI workflow components typically used by developers or technical teams.
Browser Extension AI tools installed directly in the browser. These can range from grammar checkers to screenshot tools to full AI copilot extensions and often have broad access to browser content.
IDE Plugin Developer-facing AI tools integrated into code editors, such as GitHub Copilot Chat or Claude Code for VS Code. These tools can have access to source code and repository content.
MCP Server Model Context Protocol servers that allow AI agents to interact with external services, APIs, and local data.
MCP Servers Require Extra Scrutiny
MCP Servers can grant AI models significant access to actions and data. Review every MCP Server entry carefully regardless of its assigned risk level.
Running Process AI-related processes discovered actively running in memory on the endpoint at the time of the agent scan. These may indicate AI tools currently in use even if they are not formally installed.
Background Service AI-related services configured to start automatically in the background, often at system boot, without requiring direct user interaction. These can be persistent and may go unnoticed by the device owner.
Node_env AI-related packages and libraries discovered in local Node.js project environments. Common in engineering endpoints running JavaScript-based AI agent frameworks or tooling.
Python_env AI-related packages and libraries discovered in local Python environments. Common in data science, ML, and engineering endpoints running machine learning frameworks or AI agent orchestration libraries.
Sandboxed App AI applications running in an isolated or sandboxed environment on the device. Sandboxing limits some attack surface, but these apps still warrant review for data access and governance compliance.
Filtering and Search
Use the filter controls to narrow the inventory to the assets that matter most to you:
- Risk Level — Focus on Critical or High items that require immediate attention
- Asset Type — Isolate a specific category such as MCP Servers or IDE Plugins for a targeted review
- Status — Filter to all Ungoverned assets to identify what still needs to be reviewed
- Marketplace — Filter by whether the asset is recognized in a known software marketplace
Understanding Governance Status
All newly discovered assets are marked as Ungoverned by default. This is intentional — Surface does not assume any asset is safe simply because it exists on an endpoint. Your team should review each asset and take a governance action to formally acknowledge, approve, or restrict it.
Ungoverned Does Not Mean Risky
The Ungoverned status is a prompt for action, not a risk judgment on its own. An Ungoverned Low-risk asset is a lower priority than an Ungoverned High-risk asset, but both should eventually be addressed to maintain a complete governance posture.
Using the Endpoints Count
The Endpoints column shows how many devices have a given asset installed. This number is important context when prioritizing your governance work:
- An asset rated High risk on 20 endpoints represents significant exposure and should be prioritized ahead of a High-risk asset on a single device
- An asset on 1 endpoint may indicate an individual user installed something outside of your standard software policy — worth investigating
- An asset on many endpoints that is Low risk is likely widespread legitimate software and can be approved in bulk
Recommended Workflow
- Filter by Critical and High risk levels — review and govern these first
- Filter by Ungoverned status to understand the full scope of unreviewed software
- Use the Endpoints count to prioritize assets with broad organizational exposure
- For MCP Servers, review each entry carefully regardless of risk level — their access to agent capabilities warrants extra scrutiny
Start with Risk
Always triage by risk level first. Governing your Critical and High assets delivers the most immediate reduction in your organization's AI exposure.
Dashboard
The Dashboard is your consolidated command center for AI asset activity across your organization, providing a real-time snapshot of your AI exposure, asset breakdown by type, and full risk distribution.
Endpoints
The Endpoints page shows every device enrolled in Surface monitoring, giving you visibility into the AI software footprint on any individual machine and the health of your fleet.