openclaw-veto

Veto policy checks for OpenClaw agents.

openclaw-veto intercepts every OpenClaw tool call, validates it through Veto, and either allows it, blocks it, or routes it into a human approval flow.

Install

openclaw plugins install openclaw-veto

Setup

Initialize Veto in your project so the plugin can load veto/veto.config.yaml and your rules:

cd your-project
veto init

That creates the local Veto config directory:

veto/
  veto.config.yaml
  rules/

Configure approval mode

By default, openclaw-veto uses OpenClaw's native approval UX.

Add plugin configuration to your openclaw.json:

{
  "plugins": {
    "veto": {
      "approvalMode": "openclaw-native"
    }
  }
}

Supported modes:

• openclaw-native: Veto returns require_approval, and OpenClaw handles the approval prompt with its built-in UX. This works with local YAML rules.
• veto-cloud: Veto polls Veto Cloud until the approval is resolved. Use this when you want approvals to appear in the Veto dashboard and trigger cloud workflows like notifications or webhooks.

veto-cloud requires Veto Cloud mode to be configured. Set VETO_API_KEY in the environment or configure cloud.apiKey in veto/veto.config.yaml. If Veto is only running with local YAML rules and no cloud API key, approval paths must use openclaw-native.

Example:

{
  "plugins": {
    "veto": {
      "approvalMode": "veto-cloud"
    }
  }
}

Example rules

File access

rules:
  - id: block-sensitive-files
    name: Block sensitive file reads
    enabled: true
    severity: critical
    action: block
    tools: [read_file]
    conditions:
      - field: arguments.path
        operator: matches
        value: "(^|/)(\\.env|id_rsa|id_ed25519)$"
    message: Reading secrets is not allowed

Shell commands

rules:
  - id: require-approval-for-destructive-shell
    name: Require approval for destructive shell commands
    enabled: true
    severity: high
    action: require_approval
    tools: [bash, shell]
    conditions:
      - field: arguments.command
        operator: matches
        value: "\\b(rm -rf|sudo rm|shutdown|reboot)\\b"
    message: Destructive shell commands require approval

External APIs

rules:
  - id: block-unapproved-webhooks
    name: Block outbound requests to unapproved domains
    enabled: true
    severity: high
    action: block
    tools: [http_request, fetch]
    conditions:
      - field: arguments.url
        operator: not_matches
        value: "^https://(api\\.)?(example\\.com|stripe\\.com|slack\\.com)/"
    message: Outbound API calls must target approved domains

What the plugin does

• validates every before_tool_call through veto.guard()
• blocks denied calls before execution
• supports OpenClaw-native and Veto Cloud approval flows
• records post-execution tool activity for audit visibility

Docs

For the full rule format, policy packs, cloud approvals, and deployment guidance, see https://docs.veto.so.