Policy Reference

Each rule resource has a policy attribute that accepts a JSON string (use jsonencode({...})). The structure varies by rule type. This document describes the complete schema for each.

Chats Rules (DLP Policy)

Used by: cpwai_workforce_ai_chats_rule, cpwai_browse_dlp_rule

Chats rules apply Data Loss Prevention (DLP) policies to chat/messaging events.

Workforce AI Chats Example

policy = jsonencode({
  event_type = "file_upload"
  action     = "prevent"
  logging    = "enabled"
  services_and_application = {
    mode = "selected"
    genai_application = [
      { id = 1, mode = "all" }
    ]
  }
  data_types = [
    {
      id   = "cf0523c1-537e-4a4b-8bb8-084b7b9e0b45"
      name = "Credit Card Number"
      type = "PRE_DEFINED"
    }
  ]
})

Browse Chats Example

policy = jsonencode({
  event_type = "file_upload"
  action     = "prevent"
  logging    = "enabled"
  services_and_application = {
    mode = "selected"
    domain = [
      { domain = "chat.example.com", match_mode = "any_subdomain" }
    ]
  }
  data_types = [
    {
      id   = "cf0523c1-537e-4a4b-8bb8-084b7b9e0b45"
      name = "Credit Card Number"
      type = "PRE_DEFINED"
    }
  ]
})

Fields

Field	Type	Required	Description
`event_type`	string	Yes	Event to apply DLP on. See Event Types.
`action`	string	Yes	Action when DLP detects a match. See DLP Actions.
`logging`	string	No	`"enabled"` or `"disabled"`. Default: `"enabled"`.
`services_and_application`	object	No	Scope of services. See Services and Application.
`data_types`	array	No	DLP data types to detect. See Data Types and the DLP Data Types guide.

Event Types

Value	Description
`file_upload`	File upload to a service
`file_download`	File download from a service
`paste`	Text pasted into a service
`prompt`	Text prompt sent to an AI service
`copy`	Text copied from a service

DLP Actions

Value	Description
`prevent`	Block the action entirely
`ask`	Prompt the user for confirmation
`redact`	Redact the sensitive data and allow
`detect`	Allow but log the detection
`allow`	Allow without restriction
`block`	Block the action entirely

~> Note: Rules without data_types (any-text matching) can only use block or allow actions. The detect, prevent, ask, and redact actions require at least one data type to be specified.

Access Rules

Used by: cpwai_workforce_ai_access_rule, cpwai_browse_access_rule

Access rules control which services and applications users can access.

Workforce AI Access Example

policy = jsonencode({
  action  = "block"
  logging = "enabled"
  services_and_application = {
    mode = "selected"
    genai_application = [
      {
        id   = 1
        mode = "all"
      }
    ]
  }
})

Browse Access Example

policy = jsonencode({
  action                   = "allow"
  logging                  = "enabled"
  download_file_protection = "WaitForEmulation"
  upload_file_protection   = "Detect"
  services_and_application = {
    mode = "selected"
    domain = [
      { domain = "portal.myorg.com", match_mode = "exact_host" }
    ]
  }
})

Fields

Field	Type	Required	Description
`action`	string	Yes	Action for access control: `"block"`, `"allow"`, or `"ask"`.
`logging`	string	No	`"enabled"` or `"disabled"`.
`services_and_application`	object	No	Scope of services. See Services and Application.
`download_file_protection`	string	No	File download protection action. Workforce AI rules: must be `"na"`. Browse rules: must not be `"na"` — valid values: `"Block"`, `"Allow"`, `"EmulateInBackground"`, `"WaitForEmulation"`, `"Detect"`, `"Extract"`.
`upload_file_protection`	string	No	File upload protection action. Workforce AI rules: must be `"na"`. Browse rules: must not be `"na"` — valid values: `"Block"`, `"Allow"`, `"EmulateInBackground"`, `"WaitForEmulation"`, `"Detect"`.

Agents Rules (MCP Server Policy)

Used by: cpwai_workforce_ai_agents_rule

Agents rules control AI agents and MCP (Model Context Protocol) server interactions.

policy = jsonencode({
  action  = "allow"
  logging = "enabled"
  clients = { mode = "all" }
  servers = {
    mcp_servers_mode = "manual"
    manual = [
      {
        server_type = "remote_server"
        remote_server = {
          match_mode = "domain"
          domain = {
            domain_name = "mcp.internal.com"
            match_mode  = "exact_host"
          }
        }
      }
    ]
  }
  tooling = { match_mode = "all" }
})

Fields

Field	Type	Required	Description
`action`	string	Yes	Action: `"block"`, `"allow"`, or `"ask"`.
`logging`	string	No	`"enabled"` or `"disabled"`.
`clients`	object	No	MCP client scope.
`servers`	object	No	MCP server scope.
`tooling`	object	No	Tool matching configuration.

Clients Object

Field	Type	Description
`mode`	string	`"all"` or `"selected"`
`selected`	array	List of MCP clients when `mode = "selected"`. Each: `{ os, name_id }`

Servers Object

Field	Type	Description
`mcp_servers_mode`	string	`"all"`, `"risk_level"`, `"manual"`, or `"managed"`
`risk_level`	object	Risk level configuration (when `mcp_servers_mode = "risk_level"`)
`managed`	object	Managed MCP configuration (when `mcp_servers_mode = "managed"`). Contains `name_ids` array
`manual`	array	List of MCP server definitions (when `mcp_servers_mode = "manual"`). Each has `server_type` and a type-specific config

Server types: "all", "node_server", "python_server", "claude_extensions", "executable_server", "unknown_server", "docker_gateway", "docker_runner", "remote_server".

For remote_server, the config has match_mode ("url" or "domain"), and either a domain object ({ domain_name, match_mode }) or a url object ({ url, match_mode }).

Tooling Object

Field	Type	Description
`match_mode`	string	`"tools_include"`, `"tools_exclude"`, `"operations"`, or `"all"`
`tools`	array	List of tool items. Each: `{ tool_name, match_mode = "exact" \| "contains" }`
`operations`	object	Server operations to match. Keys: `CREATE`, `READ`, `UPDATE`, `DELETE`, `OTHER`. Values: `"match"` or `"unmatch"`

Secure Browsing Rules

Used by: cpwai_browse_secure_browsing_rule

Secure browsing rules define threat-prevention policies for web browsing: phishing protection, password-reuse detection, safe search, and search reputation.

policy = jsonencode({
  password_reuse    = "prevent"
  zero_phishing     = "detect"
  safe_search       = "on"
  search_reputation = "on"
  protected_domains = [
    {
      domain     = "portal.myorg.com"
      match_mode = "exact_host"
    }
  ]
})

Fields

Field	Type	Required	Description
`password_reuse`	string	No	Password reuse detection: `"detect"`, `"prevent"`, `"ask"`, or `"off"`.
`zero_phishing`	string	No	Phishing protection: `"detect"`, `"prevent"`, `"ask"`, or `"off"`.
`safe_search`	string	No	Safe search enforcement: `"on"` or `"off"`.
`search_reputation`	string	No	Search reputation filtering: `"on"` or `"off"`.
`domain_exclusions`	array	No	Domains to exclude. Each: `{ domain, match_mode }`.
`protected_domains`	array	No	Domains to protect. Each: `{ domain, match_mode }`.

Services and Application

The services_and_application object defines which services/applications a rule applies to. Used in Chats and Access rules.

The available fields depend on whether the rule is a Workforce AI rule or a Browse rule:

Field	Workforce AI rules	Browse rules
`domain`	Not allowed	Allowed
`url`	Not allowed	Allowed
`genai_application`	Allowed	Not allowed
`category`	Only `60531762` act as any generative AI tools	Any non `60531762` category

Workforce AI Example

services_and_application = {
  mode = "selected"

  # Workforce AI Applications (by ID)
  genai_application = [
    {
      id   = 1
      mode = "all"  # or "selected_account"
      accounts = [  # only when mode = "selected_account"
        { account_id = "org-123" }
      ]
    }
  ]
}

Browse Example

services_and_application = {
  mode = "selected"

  # Domains
  domain = [
    {
      domain     = "chat.openai.com"
      match_mode = "any_subdomain"  # or "exact_host" (default)
    }
  ]

  # URLs
  url = [
    { url = "https://example.com/api" }
  ]

  # URLF Categories (by predefined ID)
  category = [
    { category_id = 1 }
  ]
}

Fields

Field	Type	Description
`mode`	string	`"all"` — apply to all services; `"selected"` — apply to listed services only
`domain`	array	Browse only. List of domains with `domain` (string) and optional `match_mode`
`url`	array	Browse only. List of URLs with `url` (string)
`genai_application`	array	GenAI only. List of Workforce AI app IDs with optional account selection. See the AI Applications Catalog guide.
`category`	array	List of URLF category IDs. Workforce AI rules: only `60531762` (Generative AI Tools) allowed. Browse rules: any category (see Supported Categories).

Domain Match Modes

Value	Description
`exact_host`	Matches the exact hostname (default)
`any_subdomain`	Matches any subdomain (e.g. `example.com` matches `sub.example.com`)

Account Selection Modes

Value	Description
`all`	Policy applies to all tool instances regardless of account (default)
`selected_account`	Policy applies only to instances connected to specified account IDs

Supported Categories

Common URLF categories that can be used in Browse rules:

Category	`category_id`
Uncategorized	`0`
Computers & Internet	`1`
Education	`3`
Entertainment	`5`
Financial Services	`7`
Gambling	`9`
Government / Military	`11`
Hacking	`13`
Health	`15`
Illegal / Questionable	`17`
Job Search / Careers	`21`
News / Media	`24`
Phishing	`31`
Pornography	`33`
Real Estate	`34`
Shopping	`42`
Business / Economy	`51`
Sports	`54`
Social Networking	`52000069`
File Storage / Sharing	`52000136`
Search Engines / Portals	`52000132`
Artificial Intelligence (AI)	`60517663`

Data Types (DLP)

Data types define the sensitive information patterns that DLP (Data Loss Prevention) rules detect in Chats rules. Each data type represents a category of sensitive data — such as credit card numbers, social security numbers, or custom patterns defined by your organization.

Using Data Types in Rules

In Chats DLP rules, reference data types in the data_types array of the policy:

data_types = [
  {
    id             = "cf0523c1-537e-4a4b-8bb8-084b7b9e0b45"
    name           = "Credit Card Number"
    type           = "PRE_DEFINED"
    matching_level = 3  # optional: sensitivity 1-5
  }
]

Field	Type	Description
`id`	string (UUID)	Unique identifier of the data type
`name`	string	Display name
`type`	string	`"PRE_DEFINED"` or `"CUSTOM"`
`matching_level`	integer	Optional. Matching sensitivity level

Use the cpwai_workforce_ai_chats_rulebase or cpwai_browse_dlp_rulebase data source to discover available predefined data types.

For the full API reference, see the DLP Datatypes API documentation.

Common Enumerations

Logging Status

Value	Description
`enabled`	Log events matching this rule
`disabled`	Do not log

Assignment Types

Value	Description
`ASSIGNMENT_TYPE_USER`	Individual user
`ASSIGNMENT_TYPE_GROUP`	User group
`ASSIGNMENT_TYPE_ENTIRE_ORG`	Entire organization

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Policy Reference

Chats Rules (DLP Policy)

Workforce AI Chats Example

Browse Chats Example

Fields

Event Types

DLP Actions

Access Rules

Workforce AI Access Example

Browse Access Example

Fields

Agents Rules (MCP Server Policy)

Fields

Clients Object

Servers Object

Tooling Object

Secure Browsing Rules

Fields

Services and Application

Workforce AI Example

Browse Example

Fields

Domain Match Modes

Account Selection Modes

Supported Categories

Data Types (DLP)

Using Data Types in Rules

Common Enumerations

Logging Status

Assignment Types

Uh oh!

FilesExpand file tree

policy-reference.md

Latest commit

History

policy-reference.md

File metadata and controls

Policy Reference

Chats Rules (DLP Policy)

Workforce AI Chats Example

Browse Chats Example

Fields

Event Types

DLP Actions

Access Rules

Workforce AI Access Example

Browse Access Example

Fields

Agents Rules (MCP Server Policy)

Fields

Clients Object

Servers Object

Tooling Object

Secure Browsing Rules

Fields

Services and Application

Workforce AI Example

Browse Example

Fields

Domain Match Modes

Account Selection Modes

Supported Categories

Data Types (DLP)

Using Data Types in Rules

Common Enumerations

Logging Status

Assignment Types