Enforce Organization-Wide MCP Policies
Compare three organization-wide MCP enforcement models across Claude Code, Cursor, and VS Code. Choose the right enforcement method for your coding assistant and organization.
JFrog enables you to enforce organization-wide MCP usage policies across your coding assistants. Rather than relying on developers to select approved servers, you define which MCP servers are available and lock that policy at the organization, team, or machine level — depending on your IDE and governance model.
Without enforcement, developers pull unverified MCP servers from public repositories, bypassing your organization's security and compliance controls. With organization-wide enforcement, you prevent shadow AI by ensuring only approved, vetted MCP servers are available to your coding agents.
Three Enforcement Models
JFrog supports three distinct enforcement models, each suited to a different coding assistant. All three prevent unapproved MCP usage — but they differ in scope, deployment method, and developer experience.
Claude Code: Managed Settings
Cloud-based policy applied organization-wide via the Claude.ai admin console, plus write protection on managed-mcp.json. The same managed settings also lock plugin marketplaces.
Cursor: Team MCP Configuration
Team-level policy applied via the Cursor team dashboard.
VS Code: MCP Allowlist Policy
Enterprise MCP allowlist (ChatAllowedMcpServers) deployed via your MDM tool (Jamf, Intune, Kandji, Mosyle). The same allowlist is applied to GitHub Copilot (allowedMcpServers) so delegated Copilot sessions cannot bypass it, and Claude and Codex delegation are disabled by default.
Compare All Three Models
| Aspect | Claude Code | Cursor | VS Code |
|---|---|---|---|
| Enforcement Model | Managed Settings | Team MCP Configuration | MCP Allowlist Policy |
| What It Governs | MCP servers and plugin marketplaces | MCP servers | MCP servers |
| Scope | Organization-wide | Team-level | Organization-wide (via MDM) |
| Where Policy Is Defined | Claude.ai admin console, plus write protection on managed-mcp.json | Cursor team dashboard | MDM enterprise policy (macOS profile or Windows script) |
Warning
VS Code reads enterprise policies only at startup. After the MCP allowlist policy is deployed via MDM, each developer must fully quit and relaunch VS Code. Reload window is not enough. Without a full restart, unapproved MCP servers may remain accessible.
Choose Your Enforcement Model
Select your coding assistant to get step-by-step enforcement instructions:
- Claude Code: Enforce with Managed Settings — Apply organization-wide policy via the Claude.ai admin console, lock the MCP allowlist and plugin marketplaces, and restrict write access to
managed-mcp.json. - Cursor: Enforce with Team MCP Configuration — Apply team-level policy via Cursor team dashboard.
- VS Code: Enforce with MCP Allowlist Policy — Deploy the enterprise allowlist via MDM (Jamf, Intune, Kandji, or Mosyle).
Frequently Asked Questions
Can I enforce MCP policies for multiple coding assistants at the same time?
Yes. If your organization uses multiple IDEs (Claude Code + Cursor, or VS Code + Claude Code), you can apply the appropriate enforcement model for each. Developers using different tools will have policies enforced according to their tool's model.
What is MDM, and why is it required for VS Code?
MDM (Mobile Device Management) is software that IT teams use to manage and deploy configurations to developer machines at scale. Examples include Jamf Pro (macOS), Microsoft Intune (Windows), and Kandji (macOS).
VS Code enforcement uses an enterprise MCP allowlist policy that must be pushed to each developer machine. MDM automates this deployment at scale. Cursor enforcement operates entirely in the cloud and doesn't require MDM. Claude Code managed settings are applied in the Claude.ai admin console, but complete enforcement also requires write protection on managed-mcp.json, which means developer accounts must not have privileged access to that file. MDM is one way to apply that restriction; it is not required. For the paths and steps, see Claude Code: Enforce Governed MCP and Plugin Usage.
What if I use VS Code but don't have MDM infrastructure?
VS Code MCP allowlist enforcement requires MDM to deploy the policy at scale. If you don't have MDM, you have two options:
- Evaluate and deploy MDM — Jamf Pro, Kandji, or Mosyle (macOS), or Microsoft Intune (Windows) are common choices. Deploying MDM is a broader security initiative that benefits multiple teams.
- Use a different IDE — If MDM is not feasible, consider Cursor team configuration, where governance doesn't require MDM. Claude Code managed settings are applied in the cloud. Completing enforcement still requires write protection on
managed-mcp.json, which means developer accounts must not have privileged access to that file. MDM is one way to apply that restriction; it is not required.
Related Topics
Updated 22 days ago