Free IDE extension for risk-free vibe coding — keep secrets out of AI.
Integration · n8n
Install n8n-nodes-soterai directly from the n8n marketplace and drag a SoterAI guard node before and after every AI step. It detects prompt injection, jailbreaks, secrets, PII, unsafe tool calls, risky memory writes, RAG poisoning, and data leakage — no command line or package install required.
Block instruction-override attempts before they reach your LLM node — including indirect injection from external data sources fetched by n8n.
Detect and redact Aadhaar, PAN, GSTIN, API keys, and other sensitive data flowing through workflow nodes before they enter the AI model.
Scan LLM responses for leaked data, unsafe content, or signs the model was steered off task before passing results to downstream nodes.
Score an exported n8n workflow for AI Agent, tool, webhook, Code node, memory, and output-egress risks before you deploy it.
The node is published in the n8n community node registry, so it installs in four steps — no npm needed.
Open n8n and go to Settings → Community Nodes
The SoterAI node is published in the n8n community node registry — no separate package install is needed.
Install n8n-nodes-soterai
Click Install a Community Node and enter n8n-nodes-soterai, or find SoterAI in the community node list.
Restart n8n if prompted
Some n8n instances require a restart after installing a community node. After it reloads, SoterAI appears in the node panel.
Search for SoterAI in the node panel
Drag the SoterAI node into any workflow. It works before or after any AI node — OpenAI, Anthropic, Ollama, or the n8n AI Agent node.
Direct node link
Create a SoterAI project and API key
Sign up at soterai.in, create a project, and generate an API key. The key starts with sk_.
Add a SoterAI API credential in n8n
In the SoterAI node, create a new SoterAI API credential and paste your API key. n8n stores it in its encrypted credential store — it never lives in workflow JSON.
Keep the default Base URL
https://soterai.in is pre-filled. Change it only when you run a self-hosted SoterAI API. Optionally set a default Project ID; each node can override it.
Run a safe test
Test with fake values first (e.g. sk-test-1234567890abcdef). Never paste real production secrets into test workflows.
API keys are handled through n8n credentials and are never stored in workflow JSON. Use fake test values such as sk-test-1234567890abcdef in demos and screenshots.
| Operation | Purpose |
|---|---|
| Universal AI Firewall (Best Protection) | Recommended one-node protection. Checks prompt injection, jailbreaks, PII/secrets, RAG context, tool calls, memory operations, AI output, and semantic data egress in a single node. |
| Guard Input | Check inbound prompts before an AI app receives them. Supports Block, Redact, Warn, or Continue. |
| Guard Output | Check AI-generated output before sending, saving, or responding with it. Supports Block, Redact, Warn, or Continue. |
| Redact Secrets or PII | Detect and redact sensitive strings such as emails, phone numbers, API keys, and secrets. |
| Get RAG Risk Summary | Scan a document or chunk and return trustScore, trustLevel, findings, and a recommended action. |
| Audit n8n Workflow Security | Score an exported n8n workflow for AI Agent, tool, webhook, Code node, memory, RAG, credential, and output-egress risks. |
| Analyze Text | Analyze a text field and return allowed, riskScore, categories, reason, and safe text without blocking. |
Choose Universal AI Firewall (Best Protection) for the simplest and strongest workflow pattern:
User/Webhook Input → SoterAI Universal AI Firewall (Input Text = user message) → LLM → SoterAI Universal AI Firewall (AI Output Text = model response) → Respond / Tool / Memory
Fill Input Text with the incoming message, add AI Output Text (Optional) to scan the model response, and keep Maximum Protection with On Threat = Block for production flows. Optional layers for RAG context, tool calls, memory operations, and output destinations live under Security Context.
Every SoterAI node has two outputs and routes items itself — you do not need an IF node to act on a verdict.
┌─ Safe ────► rest of your workflow
Webhook ──► SoterAI ─┤
└─ Flagged ─► respond "blocked", log, or leave unconnectedThe node ships with importable example workflows. Import any file from examples/ in the package.
Manual Trigger → SoterAI Analyze Text → IF High Risk.
Webhook → SoterAI Guard Input → IF Risk High → Respond to Webhook.
Manual Trigger → AI Output Text → SoterAI Guard Output → Save Safe Output.
Manual Trigger → SoterAI Redact Secrets or PII → IF Secrets Found → Safe Output.
Webhook → SoterAI Universal AI Firewall → blocked / allowed response branches.
Manual Trigger → SoterAI Audit n8n Workflow Security → posture report.
No. The node is published in the n8n community node registry, so you install it directly from n8n (Settings → Community Nodes) without touching the command line.
n8n-nodes-soterai is published on the n8n community node registry and npm. The direct node page is https://www.npmjs.com/package/n8n-nodes-soterai.
SoterAI is a standalone community node with Safe and Flagged outputs. It works before or after any AI node — OpenAI, Anthropic, Ollama, or custom LLM nodes — as well as with n8n's AI Agent node.
SoterAI adds under 100ms per guard check. For most n8n workflows this is negligible compared to the LLM API call itself (typically 1-10 seconds).
Yes. The node also supports a custom Base URL, so both SoterAI and n8n can run on the same self-hosted infrastructure, keeping all data within your network boundary.
The node routes items itself. Safe items go to the Safe output and stopped items go to the Flagged output — no IF node needed. In Monitor-style settings it logs and continues so you can review patterns without blocking production workflows.
Install the node from n8n Settings → Community Nodes, create a free SoterAI project, and guard your first workflow in under 10 minutes.