>
    Strata Copilot
    Set Up the Enterprise DLP MCP Server
    Focus
    Download PDF
    Focus
    1. Home
    2. Enterprise DLP
    3. Administration
    4. Configure Enterprise DLP
    5. Enterprise DLP MCP Server
    6. Set Up the Enterprise DLP MCP Server
    Download PDF
    Enterprise DLP

    Set Up the Enterprise DLP MCP Server

    Table of Contents

    Set Up the Enterprise DLP MCP Server

    Connect your AI agent to Enterprise DLP by letting it configure the MCP server from a generated prompt, or by configuring the server yourself.
    Where Can I Use This?What Do I Need?
    • NGFW (Managed by Panorama or Strata Cloud Manager)
    • Prisma Access (Managed by Panorama or Strata Cloud Manager)
    • SaaS Security Inline
    • Data Security
    • Prisma AIRS
    • Prisma Browser
    • Email DLP
    • Endpoint DLP
    • Enterprise Data Loss Prevention (E-DLP) license, or one of the following licenses that include it:
      • Prisma Access CASB license
      • Next-Generation CASB for Prisma Access and NGFW (CASB-X) license
      • Data Security license
    • An AI agent and AI model that support the Model Context Protocol (MCP)
    • Node.js 18 or later installed on the machine that runs your AI agent
    • A service account with a role that grants view access to Enterprise DLP
    Use the MCP server to view your Enterprise DLP configurations and data regardless of the enforcement point where they are applied.
    Set up the Enterprise Data Loss Prevention (E-DLP) MCP server to connect your AI agent to your Enterprise DLP tenant so the agent can view, and optionally change, your Enterprise DLP incidents and configuration in natural language. Setup has three parts: you create a dedicated service account that gives your agent a limited identity, you choose whether the agent has read-only or read-write access, and you register the MCP server in your agent. The MCP server runs locally from the public @paloaltonetworks/mcp-dlp npm package using npx, so you don't download a separate application. The Enterprise DLP describes the components of the integration and how they work together. The MCP server supports any MCP-compatible AI agent. The procedures in this topic cover the most popular agents: Claude Desktop, Claude Code, OpenAI Codex, Cursor, and VS Code with GitHub Copilot. If you use a different agent, follow its documentation to add an MCP server.
    You can set up the MCP server in two ways:
    • Prompt—Strata Cloud Manager generates a prompt that you paste into your agent. The agent verifies that Node.js 18 or later is installed, sets the environment variables, configures the MCP server for your agent, and runs a live verification, so you don't edit configuration files by hand. This is the quickest method.
    • Manual—You test your service account credentials from the command line and add the MCP server to your agent's configuration yourself. Use this method if your organization prevents your agent from registering MCP servers, requires change control for agent configuration, or if you prefer to configure the connection yourself.
    Important Considerations
    • Service account—Palo Alto Networks recommends a dedicated service account for each MCP server integration rather than sharing an existing admin account. A dedicated account limits what the agent can do, lets you revoke its access without affecting other automation, and gives you a clear audit trail of agent activity. Don't use the same service account for multiple MCP server integrations.
    • Role—You can assign a predefined or custom role to the service account. The role limits what your agent can do, regardless of the access level you choose.
    • Access level—Palo Alto Networks recommends read-only access. With read-only access, your agent can view incidents, data profiles, data patterns, and reports but can't change anything, and the MCP server doesn't make write or delete operations available, even if the service account has broader permissions. With read-write access, your agent can also create, update, and delete Enterprise DLP configurations and incidents.
    • CLI fallback—If your organization prevents your agent from registering MCP servers, you can ask your agent to use the command-line interface (CLI) instead. The CLI reads and writes to your tenant as the MCP server does, but you can't use MCP server features such as a predefined skill.

    Prompt

    Set up the Enterprise DLP MCP server using a prompt that Strata Cloud Manager generates.
    1. Install an AI agent.
      You need an AI agent that supports MCP to connect to Enterprise DLP. If you use an agent that isn't listed, follow its installation instructions. Otherwise, install one of the following popular agents if you haven't already:
      • Claude Desktop
        Anthropic's desktop app with built-in MCP support for extensions.
        1. Download Claude Desktop from claude.ai/download.
        2. (macOS) Open the .dmg file and drag Claude to the Applications folder.
        3. (Windows) Run the .exe installer and follow the prompts.
        4. Sign in with your Anthropic account.
      • Claude Code
        Anthropic's terminal-based coding agent, ideal for developers.
        1. Install Claude Code with npm: npm install -g @anthropic-ai/claude-code
        2. (macOS) As an alternative to npm, install Claude Code with Homebrew: brew install claude-code
        3. Launch Claude Code by running claude.
        4. When prompted, sign in with your Anthropic account.
      • OpenAI Codex
        OpenAI's coding agent with MCP support through the command-line interface (CLI).
        1. Install OpenAI Codex with npm: npm install -g @openai/codex
        2. Launch OpenAI Codex by running codex.
        3. When prompted, sign in with your OpenAI account.
      • Cursor
        An AI-native code editor with built-in MCP support.
        1. Download Cursor from cursor.sh.
        2. Install and open the application.
        3. Cursor supports MCP natively, so you configure MCP servers in your project settings.
      • VS Code with GitHub Copilot
        Visual Studio Code with GitHub Copilot agent mode.
        1. Download VS Code from code.visualstudio.com.
        2. Install the GitHub Copilot extension from the Extensions marketplace.
        3. Enable Copilot agent mode in the VS Code settings.
    2. Log in to Strata Cloud Manager.
    3. Create a new service account for your AI agent.
      1. Select System SettingsIdentity & AccessAccess Management and Add Identity.
      2. For Identity Type, select Service Account.
      3. Click Download CSV File to download the service account Client ID and Client Secret as a CSV file.
        The agent uses this CSV file during the initial setup.
      4. Assign a role to the service account.
        You can assign a predefined or custom role to your agent service account.
        • For Apps & Services, select Enterprise DLP.
        • For Role, select the predefined or custom role with the access privileges you want to assign to the agent.
      5. Submit.
    4. Select ConfigurationData Loss PreventionSettingsMCP Server.
    5. In Choose Access Level, select the access level for your agent.
      Your selection sets the environment variable for the MCP server in the generated prompt. Palo Alto Networks recommends read-only access.
    6. Click Copy this Prompt into Your Agent.
      You can expand View prompt to review the prompt first.
    7. Paste the prompt into your AI agent.
      Your agent verifies that Node.js 18 or later is installed, sets the environment variables, configures the MCP server for your agent, and runs a live verification. The MCP server runs through npx from the public @paloaltonetworks/mcp-dlp npm package, so you do not download anything.
      If the agent doesn't find Node.js, it prompts you to confirm that you want to install it. You can install Node.js in the default location or you can install it in any directory that you require
      The default installation location is:
      • Windows—C:\Program Files\nodejs\
      • macOS/Linux—/usr/local/bin/
    8. When your agent prompts you, confirm that you downloaded the agent service account CSV, and paste your tenant TSG ID.
    9. Wait for your AI agent to complete the MCP server configuration.
      The agent then performs a live check by retrieving the last 7 days of DLP incidents to validate the connection.
      The AI agent can't register the Enterprise DLP MCP server if your organization prevents the creation of MCP servers.
      If you receive an MCP registration blocked error, prompt the AI agent to use the CLI instead of the MCP server.
      Instructing the agent to use the CLI instead still allows you to read and write to your tenant as you would with the MCP server. However, you can't use MCP server-specific features such as a predefined skill.
    10. Confirm that your agent can successfully reach your Enterprise DLP tenant.
      Ask your agent to list all of my data patterns. If your agent returns your configured data patterns, it can read your Enterprise DLP tenant.

    Manual

    Set up the Enterprise DLP MCP server manually by testing your credentials and adding the server to your AI agent's configuration.
    1. Install an AI agent.
      You need an AI agent that supports MCP to connect to Enterprise DLP. If you use an agent that isn't listed, follow its installation instructions. Otherwise, install one of the following popular agents if you haven't already:
      • Claude Desktop
        Anthropic's desktop app with built-in MCP support for extensions.
        1. Download Claude Desktop from claude.ai/download.
        2. (macOS) Open the .dmg file and drag Claude to the Applications folder.
        3. (Windows) Run the .exe installer and follow the prompts.
        4. Sign in with your Anthropic account.
      • Claude Code
        Anthropic's terminal-based coding agent, ideal for developers.
        1. Install Claude Code with npm: npm install -g @anthropic-ai/claude-code
        2. (macOS) As an alternative to npm, install Claude Code with Homebrew: brew install claude-code
        3. Launch Claude Code by running claude.
        4. When prompted, sign in with your Anthropic account.
      • OpenAI Codex
        OpenAI's coding agent with MCP support through the command-line interface (CLI).
        1. Install OpenAI Codex with npm: npm install -g @openai/codex
        2. Launch OpenAI Codex by running codex.
        3. When prompted, sign in with your OpenAI account.
      • Cursor
        An AI-native code editor with built-in MCP support.
        1. Download Cursor from cursor.sh.
        2. Install and open the application.
        3. Cursor supports MCP natively, so you configure MCP servers in your project settings.
      • VS Code with GitHub Copilot
        Visual Studio Code with GitHub Copilot agent mode.
        1. Download VS Code from code.visualstudio.com.
        2. Install the GitHub Copilot extension from the Extensions marketplace.
        3. Enable Copilot agent mode in the VS Code settings.
    2. Log in to Strata Cloud Manager.
    3. Create a new service account for your AI agent.
      1. Select System SettingsIdentity & AccessAccess Management and Add Identity.
      2. For Identity Type, select Service Account.
      3. Click Download CSV File to download the service account Client ID and Client Secret as a CSV file.
        You enter the Client ID and Client Secret as environment variables later in this procedure.
      4. Assign a role to the service account.
        You can assign a predefined or custom role to your agent service account.
        • For Apps & Services, select Enterprise DLP.
        • For Role, select the predefined or custom role with the access privileges you want to assign to the agent.
      5. Submit.
    4. Decide the access level for your agent.
      For read-only access, which Palo Alto Networks recommends, you set the DLP_READ_ONLY environment variable to 1. With read-only access, your agent can view incidents, data profiles, data patterns, and reports but cannot change anything, even if the service account has broader permissions.
    5. Verify that Node.js 18 or later is installed on the machine that runs your agent.
      The MCP server runs through npx from the public @paloaltonetworks/mcp-dlp npm package, so you do not download anything else.
      Open a terminal and run node --version to verify the Node.js version.
      If you need to install Node.js, the default installation location is:
      • Windows—C:\Program Files\nodejs\
      • macOS/Linux—/usr/local/bin/
      However, you can install it in any directory that you require.
    6. (Optional) Install the MCP server package globally.
      By default, your agent runs the MCP server through npx, which downloads the package when the agent starts the server. To install the package on your machine instead, run the following command in any terminal, such as Command Prompt or PowerShell on Windows, or Terminal on macOS and Linux:
      npm install -g @paloaltonetworks/mcp-dlp@latest
      If you receive an EACCES permissions error on macOS or Linux, follow npm's guidance on resolving permissions errors.
    7. Set your service account credentials as environment variables in your shell.
      Set DLP_CLIENT_ID, DLP_CLIENT_SECRET, DLP_TSG_ID, and DLP_READ_ONLY. For example, in a macOS or Linux shell:
      In Windows PowerShell:
      $env:DLP_CLIENT_ID="<client_id>"
$env:DLP_CLIENT_SECRET="<client_secret>"
$env:DLP_TSG_ID="<tsg_id>"
$env:DLP_READ_ONLY="1"
      • macOS or Linux
        Add the following to ~/.zshrc or ~/.bashrc:
        export DLP_CLIENT_ID="<client_id>"
export DLP_CLIENT_SECRET="<client_secret>"
export DLP_TSG_ID="<tsg_id>"
(Read-only access) export DLP_READ_ONLY=1
source ~/.zshrc or source ~/.bashrc
      • Windows PowerShell
        Run the following commands in Windows PowerShell.
        $env:DLP_CLIENT_ID="<client_id>"
$env:DLP_CLIENT_SECRET="<client_secret>"
$env:DLP_TSG_ID="<tsg_id>"
(Read-only access) $env:DLP_READ_ONLY="1"
    8. Install the MCP server.
      The MCP server runs through npx from the public @paloaltonetworks/mcp-dlp npm package, so you do not download any additional binary.
      1. Open any terminal and verify that Node.js 18 or later is installed on the machine that runs your agent.
        node --version
      2. Test that the npm package runs successfully.
        npx -y @paloaltonetworks/mcp-dlp cli --help
      3. Make a live API call to confirm the credentials from the previous step work as expected.
        npx -y @paloaltonetworks/mcp-dlp cli incidents list --page-size 3
    9. Add the MCP server to your AI agent's MCP configuration.
      Replace <your-client-id>, <your-client-secret>, and <your-tsg-id> with your service account credentials. To use read-only access, also add the DLP_READ_ONLY variable with a value of 1 alongside the other variables.
      • Claude Desktop and Claude Code
        Add the following to ~/.claude_desktop_config.json for Claude Desktop or ~/.claude.json for Claude Code:
        {
  "mcpServers": {
    "paloaltonetworks-dlp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@paloaltonetworks/mcp-dlp"
      ],
      "env": {
        "DLP_CLIENT_ID": "<your-client-id>",
        "DLP_CLIENT_SECRET": "<your-client-secret>",
        "DLP_TSG_ID": "<your-tsg-id>"
      }
    }
  }
}
      • Cursor
        Add the following to .cursor/mcp.json:
        {
  "mcpServers": {
    "paloaltonetworks-dlp": {
      "command": "npx",
      "args": [
        "-y",
        "@paloaltonetworks/mcp-dlp"
      ],
      "env": {
        "DLP_CLIENT_ID": "<your-client-id>",
        "DLP_CLIENT_SECRET": "<your-client-secret>",
        "DLP_TSG_ID": "<your-tsg-id>"
      }
    }
  }
}
      • VS Code
        Add the following to .vscode/mcp.json:
        {
  "servers": {
    "paloaltonetworks-dlp": {
      "command": "npx",
      "args": [
        "-y",
        "@paloaltonetworks/mcp-dlp"
      ],
      "env": {
        "DLP_CLIENT_ID": "<your-client-id>",
        "DLP_CLIENT_SECRET": "<your-client-secret>",
        "DLP_TSG_ID": "<your-tsg-id>"
      }
    }
  }
}
      • OpenAI Codex
        Add the following to ~/.codex/config.toml:
        [mcp_servers.paloaltonetworks-dlp]
command = "npx"
args = ["-y", "@paloaltonetworks/mcp-dlp"]

[mcp_servers.paloaltonetworks-dlp.env]
DLP_CLIENT_ID = "<your-client-id>"
DLP_CLIENT_SECRET = "<your-client-secret>"
DLP_TSG_ID = "<your-tsg-id>"
      • Other AI agents
        Add a standard I/O (stdio) MCP server to your agent's configuration with the following values. The file and format depend on your agent, so follow its documentation.
        • Name—paloaltonetworks-dlp
        • Command—npx
        • Arguments—-y and @paloaltonetworks/mcp-dlp
        • Environment variables—DLP_CLIENT_ID, DLP_CLIENT_SECRET, and DLP_TSG_ID
      Don't store your Client Secret in a configuration file that you commit to a repository or share. Keep it in your environment or in your agent's protected settings.
    10. Restart your AI agent so that it loads the MCP server.
    11. Confirm that your agent can read your Enterprise DLP tenant.
      Ask your agent to list all of my data patterns. If your agent returns your configured data patterns, the MCP server works. If your organization still prevents your agent from using the MCP server, you can ask your agent to use the CLI instead. The CLI reads and writes to your tenant in the same way, but you can't use MCP-specific features such as a predefined skill.
      The AI agent can't register the Enterprise DLP MCP server if your organization prevents the creation of MCP servers.
      If you receive an MCP registration blocked error or are otherwise unable to use the MCP server to read from and write to your Enterprise DLP tenant, prompt the AI agent to use the CLI instead of the MCP server.
      Instructing the agent to use the CLI instead still allows you to read and write to your tenant as you would with the MCP server. However, you can't use MCP server-specific features such as a predefined skill.

    © 2026 Palo Alto Networks, Inc. All rights reserved.