n8n Workflow Scanner

AI-BOM is the first security tool to scan n8n workflows for AI components. The n8n scanner detects AI Agent nodes, MCP client connections, hardcoded credentials, and other security risks in n8n workflow JSON files.

Scanning methods

Scan workflow files

# Scan exported workflow JSON files
ai-bom scan ./workflows/

# Scan a specific workflow file
ai-bom scan workflow.json

Scan local n8n installation

ai-bom scan . --n8n-local

This scans the ~/.n8n/ directory where n8n stores workflow data.

Scan running n8n instance via API

ai-bom scan . --n8n-url http://localhost:5678 --n8n-api-key YOUR_KEY

This connects to a running n8n instance and scans all workflows via the n8n API.

What it detects

AI node types

The scanner identifies these n8n AI node types:

Node Type	Category
@n8n/n8n-nodes-langchain.agent	AI Agents
@n8n/n8n-nodes-langchain.lmChatOpenAi	LLM Chat (OpenAI)
@n8n/n8n-nodes-langchain.lmChatAnthropic	LLM Chat (Anthropic)
@n8n/n8n-nodes-langchain.lmChatGoogleGemini	LLM Chat (Google)
@n8n/n8n-nodes-langchain.mcpClientTool	MCP Client
@n8n/n8n-nodes-langchain.toolCode	Code Tools
@n8n/n8n-nodes-langchain.toolWorkflow	Workflow Tools
@n8n/n8n-nodes-langchain.embeddingsOpenAi	Embeddings
@n8n/n8n-nodes-langchain.vectorStoreInMemory	Vector Stores
@n8n/n8n-nodes-langchain.chainSummarization	Summarization Chains

Security risks

• Hardcoded credentials - API keys embedded directly in workflow JSON parameters
• Webhook triggers without authentication - Workflows exposed without auth headers
• Dangerous tool combinations - Code execution tools connected to AI agents without guardrails
• Cross-workflow AI chains - Multi-workflow AI agent architectures

n8n community node

For scanning directly from the n8n UI, install the Trusera community node:

npm install n8n-nodes-trusera

Or via the n8n UI: Settings > Community Nodes > Install > n8n-nodes-trusera

Setup

1. Add a Trusera Dashboard node to a workflow
2. Create credentials with your n8n API URL and API key
3. Optionally set a dashboard password for AES-256-GCM encryption
4. Execute the node - it fetches all workflows, scans them, and returns an interactive HTML dashboard

Dashboard features

• Severity distribution charts and risk score stat cards
• Sortable findings table with search and severity/type filters
• Per-finding remediation cards with fix steps and guardrail recommendations
• OWASP LLM Top 10 category mapping for every risk flag
• CSV and JSON export
• Light/dark theme toggle
• Optional password protection (AES-256-GCM encrypted, client-side decryption)

Example output

Scanning an n8n workflow with AI agents produces output like:

Component: n8n AI Agent
  Type: agent
  Source: workflow.json (node: "AI Research Agent")
  Risk Score: 65 (HIGH)
  Flags: ai_agent, tool_use, no_guardrails
  Workflow: "Research Pipeline" (id: abc123)

Component: OpenAI LLM Chat
  Type: llm_provider
  Source: workflow.json (node: "OpenAI Chat Model")
  Risk Score: 45 (MEDIUM)
  Flags: llm_provider, model_reference
  Model: gpt-4o

Risk scoring for n8n

n8n components are scored based on:

• Node type - AI agents score higher than simple LLM calls
• Tool connections - Agents with code execution tools score higher
• Authentication - Missing webhook auth increases the score
• Credentials - Hardcoded credentials are critical severity
• Agent chains - Multi-agent workflows score higher due to complexity