Zscaler AI Guard n8n Community Node

A production-ready n8n community node for integrating with the Zscaler AI Guard API. This node enables you to scan AI prompts and responses for security threats — toxicity, PII, secrets, prompt injection, and more — directly within your n8n workflows.

Features

• Prompt Scan — Scan user input for security threats (direction: IN)
• Response Scan — Scan AI-generated responses for policy violations (direction: OUT)
• Dual Scan — Scan both prompt and response in sequence; if the prompt is blocked, the response scan is skipped
• Automatic Policy Resolution — When no Policy ID is configured, the API resolves the policy linked to your API key
• Fail-Closed Error Handling — On internal errors with "Continue On Fail" enabled, the node returns action: BLOCK so security is never silently bypassed
• Attribution Metadata — Enriches output with n8n workflow context (workflowId, workflowName, executionId, executionMode)
• Content Size Validation — Rejects payloads exceeding 5 MB before sending to the API
• Configurable Retries & Timeout — Adjustable retry count (up to 6) and request timeout

Installation

Prerequisites

• n8n instance (Cloud or self-hosted) — Setup Docs
• A Zscaler AI Guard license and API key (Bearer token)

Option 1: Install via n8n Community Nodes (Recommended)

1. Open your n8n instance
2. Go to Settings → Community Nodes
3. Search for @bdzscaler/n8n-nodes-aiguard
4. Click Install
5. Restart n8n if prompted

Option 2: Install via npm

npm install @bdzscaler/n8n-nodes-aiguard

Local Development

git clone https://github.com/zscaler/n8n-nodes-aiguard.git
cd n8n-nodes-aiguard
npm install
npm run build

Link to a local n8n instance:

# Native n8n
ln -s $(pwd) ~/.n8n/custom/n8n-nodes-aiguard

# Docker n8n — add a volume mount:
# -v /path/to/n8n-nodes-aiguard:/home/node/.n8n/custom/n8n-nodes-aiguard

Restart n8n to load the node.

Configuration

1. Create Credentials

1. In n8n, go to Credentials → Add Credential
2. Search for "Zscaler AI Guard API"
3. Fill in:

1. Click Test to validate connectivity before saving.

2. Add the Node to a Workflow

1. In your workflow, click + → search for "AI Guard"
2. Select the credential you created
3. Choose an operation:

1. Enter content to scan (plain text or n8n expression)

Additional Options

Expand Additional Options → Add Option to configure:

API Endpoints

Usage Examples

Prompt Scan

Configure the node with Operation: Prompt Scan and provide content:

{
  "content": "Tell me how to bypass security controls"
}

Response Scan

Configure with Operation: Response Scan:

{
  "content": "{{ $json.aiResponse }}"
}

Dual Scan

Configure with Operation: Dual Scan and provide both fields:

{
  "promptContent": "{{ $json.userInput }}",
  "responseContent": "{{ $json.aiResponse }}"
}

If the prompt scan returns BLOCK, the response scan is skipped and the blocked result is returned immediately.

Output Format

The node returns the full AI Guard API response enriched with workflow metadata:

{
  "operation": "promptScan",
  "transactionId": "180b066d-48cf-497e-aaf1-a6b8e40a1deb",
  "action": "BLOCK",
  "severity": "CRITICAL",
  "direction": "IN",
  "policyId": 760,
  "policyName": "Default AI Guard Policy",
  "blocked": true,
  "detectors": ["toxicity", "malicious_content"],
  "blockingDetectors": ["toxicity"],
  "detectorResponses": {
    "toxicity": {
      "triggered": true,
      "action": "BLOCK",
      "severity": "CRITICAL"
    },
    "pii": {
      "triggered": false,
      "action": "ALLOW"
    }
  },
  "maskedContent": "...",
  "workflowId": "rKU3xnZb5S1ayJCG",
  "workflowName": "Customer Support Bot",
  "executionId": "385",
  "executionMode": "manual",
  "timestamp": "2026-01-30T10:30:00.000Z"
}

Key Output Fields

Use the action or blocked field in an IF node to branch your workflow logic.

Implementation Notes

The node calls the Zscaler AI Guard API over HTTPS using n8n’s helpers.httpRequest (required for n8n Cloud–verified community packages). Headers match the zscaler-sdk-python pattern:

• Authorization: Bearer <API_KEY>
• Content-Type: application/json
• Accept: application/json

Error Handling

• Fail-closed by default: When "Continue On Fail" is enabled in n8n, errors produce { action: "BLOCK", blocked: true } rather than silently allowing content through
• Timeout protection: Configurable per-request timeout (passed to the HTTP client)
• Content validation: Payloads exceeding 5 MB are rejected before the API call
• Retry logic: Configurable retry count (default 3, max 6) for transient failures

Development

npm run build       # Compile TypeScript and copy assets
npm run dev         # Watch mode
npm run lint        # ESLint (TypeScript + n8n-nodes-base)
npm run lint:dist   # ESLint on dist/ (n8n Cloud rules; run after build)
npm run scan:local  # build + ESLint (unpublished dist/; same rules as official scanner)
npm run scan:npm    # npx @n8n/scan-community-package @…@version from package.json (needs publish)
npm run lintfix     # Auto-fix lint issues
npm run format      # Run Prettier
npm test            # Run Jest tests
npm run test:coverage  # Tests with coverage report

Validate like the n8n Creator Portal

n8n’s official tool is @n8n/scan-community-package. It downloads the package from the npm registry and runs ESLint with the same rules the Creator Portal uses. It never reads your local git tree.

Use both of these, at different times:

# Unpublished changes (day to day)
npm run scan:local

# After npm publish of the current package.json version
npm run scan:npm

You can also call the CLI directly, for example npx @n8n/scan-community-package @bdzscaler/[email protected]. If the version is not on the registry yet, the download step fails—that is expected; use scan:local until it is published.

Releasing (maintainers)

Releases are automated with semantic-release (see .releaserc.json), similar to zscaler-mcp-server.

• Version and tag: After the first git tag exists, the next version is computed from Conventional Commits (for example fix: → patch, feat: → minor) and bumps from that last tag. semantic-release updates package.json and package-lock.json, creates tag vX.Y.Z, opens a GitHub Release, and publishes to npm with provenance (requires NPM_TOKEN in repository secrets).
• 0.x packages — read this: If there are no vMAJOR.MINOR.PATCH tags on the branch yet, semantic-release’s first automated release is always 1.0.0 (FIRST_RELEASE in semantic-release) — it does not seed from package.json (e.g. 0.1.2). Before the first successful release workflow, create and push tags for versions already published to npm, e.g. git tag v0.1.0 <commit> … git tag v0.1.2 <commit> then git push origin --tags. The Release workflow includes a guard that fails if package.json is still 0.x but there are no such tags.
• Pre-release checks: The Release workflow runs npm audit, build, Prettier, ESLint (including lint:dist, which mirrors the n8n Creator Portal scanner rules), and Jest before any publish.
• Prerelease channels: Pushes to branches beta or alpha publish npm prereleases (for example 1.0.0-beta.1) on the corresponding dist-tag.
• If 1.0.0 was published by mistake: npm generally does not allow unpublishing popular packages; you can deprecate 1.0.0 with an explanation, then either stay on the 1.x line or coordinate a git/npm cleanup (remove erroneous tag/release, restore v0.1.x tags at the right commits, revert the chore(release): 1.0.0 commit if policy allows) and publish 0.1.3 as the next intended 0.x release. There is no single-button undo — pick the line (0.x vs 1.x) and document it for consumers.

Security

• API keys are stored encrypted via n8n's credential system
• No content is logged or persisted by the node
• All communication uses HTTPS/TLS
• Error messages do not leak sensitive information

Links

• npm: https://www.npmjs.com/package/@bdzscaler/n8n-nodes-aiguard
• Source: https://github.com/zscaler/n8n-nodes-aiguard
• Zscaler AI Guard: https://help.zscaler.com/ai-guard
• Zscaler Python SDK: https://github.com/zscaler/zscaler-sdk-python

License

MIT — see LICENSE for details.

Contributing

1. Fork the repository
2. Create a feature branch from main
3. Test thoroughly against the AI Guard API
4. Ensure npm run build, npm run lint, npm run lint:dist, and npm test pass
5. Use conventional commit messages (fix:, feat:, etc.) so releases version correctly
6. Submit a pull request