Vitaliy This workflow corresponds to n8n.io template #12725 — we link there as the canonical source.
The workflow JSON
Copy or download the full n8n JSON below. Paste it into a new n8n workflow, add your credentials, activate. Full import guide →
{
"id": "zbpHuVvtlRSOE27NGyyGs",
"name": "Verify AI draft answers with Pearl Hybrid Intelligence and Open AI",
"tags": [],
"nodes": [
{
"id": "05107af8-a4e3-4aa0-a1e9-f0ce0ae57148",
"name": "Overview",
"type": "n8n-nodes-base.stickyNote",
"position": [
-544,
48
],
"parameters": {
"width": 528,
"height": 540,
"content": "# Verify AI answers with Pearl Hybrid Intelligence and Open AI\n\n## How it works\n1. A user asks a question.\n2. If anything is missing, the assistant asks one focused follow-up question.\n3. Once ready, the assistant drafts an answer with OpenAI.\n4. The draft is reviewed by a licensed Pearl expert.\n5. The user receives the verified answer with expert attribution.\n\n## Setup steps\n1. Add your OpenAI credential in n8n.\n2. Add your Pearl MCP Server API key (request a demo key at: https://www.pearl.com/enterprise/contact-get-started).\n3. Activate the workflow and call the webhook with your conversation messages.\n4. Adjust the intake + drafting prompts in **Prepare Input** to match your domain and tone."
},
"typeVersion": 1
},
{
"id": "f37c8600-de53-40e3-8a2d-0a2d58158860",
"name": "Pearl Verify Webhook",
"type": "n8n-nodes-base.webhook",
"position": [
160,
368
],
"parameters": {
"path": "pearl-verify",
"options": {},
"httpMethod": "POST",
"responseMode": "responseNode"
},
"typeVersion": 2.1
},
{
"id": "18b2c188-327d-4823-972f-7f0d7f38cf8e",
"name": "Prepare Input",
"type": "n8n-nodes-base.code",
"position": [
384,
368
],
"parameters": {
"jsCode": "/**\n * n8n Code node (JavaScript)\n * Purpose:\n * - Accept webhook body containing an OpenAI Chat Completions-style `messages` array:\n * [{ role: \"user\"|\"assistant\"|\"system\", content: \"...\" }, ...]\n * - Construct two message arrays:\n * 1) messagesClarify: system prompt + messages (forces output: END or 1 clarifying question)\n * 2) messagesDraft: system prompt + messages (produces draft answer)\n */\n\nfunction assertValidMessages(messages) {\n if (!Array.isArray(messages)) {\n throw new Error('Request body must include a \"messages\" array.');\n }\n\n for (let i = 0; i < messages.length; i++) {\n const m = messages[i];\n if (typeof m !== \"object\" || m === null) {\n throw new Error(`messages[${i}] must be an object.`);\n }\n if (typeof m.role !== \"string\" || !m.role.trim()) {\n throw new Error(`messages[${i}].role must be a non-empty string.`);\n }\n if (typeof m.content !== \"string\" || !m.content.trim()) {\n throw new Error(`messages[${i}].content must be a non-empty string.`);\n }\n }\n}\n\nfunction getLastUserMessage(messages) {\n for (let i = messages.length - 1; i >= 0; i--) {\n if (messages[i].role === \"user\") return messages[i].content;\n }\n return \"\";\n}\n\n// n8n webhook payload can appear as items[0].json.body (Webhook node) or items[0].json (some setups)\nconst input = $input.first() ?? {};\nconst body = input.json.body ?? input;\n\n// Read request fields\nconst messages = body.messages;\nconst openaiModel = body.model || \"gpt-4o-mini\";\n\n// Validate\nassertValidMessages(messages);\n\n// Optional: strip any accidental leading/trailing whitespace\nconst normalizedMessages = messages.map((m) => ({\n role: m.role.trim(),\n content: m.content.trim(),\n}));\n\nconst userQuestion = getLastUserMessage(normalizedMessages);\n\n// System prompt: Clarification gate (must output END or a single question)\nconst clarificationSystem = {\n role: \"system\",\n content: [\n \"You are an intake assistant for complex legal and health questions.\",\n \"\",\n \"Your job is NOT to answer yet. First decide if you need more context.\",\n \"\",\n \"Rules:\",\n \"1) If you need more information to answer safely/accurately, respond with ONLY the single best clarifying question.\",\n \" - No preamble, no bullet points, no multiple questions.\",\n \"2) If you do NOT need any clarification, respond with EXACTLY this single token: END\",\n \"\",\n \"Output must be plain text only.\"\n ].join(\"\\n\"),\n};\n\n// System prompt: Draft answer generation (used only when clarification returns END)\nconst draftSystem = {\n role: \"system\",\n content: [\n \"You are a helpful assistant.\",\n \"\",\n \"Provide a draft answer to the user's question using the full conversation context.\",\n \"\",\n \"IMPORTANT SAFETY/QUALITY REQUIREMENTS (health/legal):\",\n \"- Provide general informational guidance only; do NOT claim to be a professional.\",\n \"- Encourage consulting a licensed professional for decisions.\",\n \"- If health: if there are urgent or dangerous symptoms, advise urgent care/emergency services.\",\n \"- If legal: mention jurisdiction matters and advise consulting a qualified attorney.\",\n \"\",\n \"Structure:\",\n \"1) Short summary\",\n \"2) Key considerations (bullets)\",\n \"3) Next steps\"\n ].join(\"\\n\"),\n};\n\n// Build the two request message arrays\nconst messagesClarify = [clarificationSystem, ...normalizedMessages];\nconst messagesDraft = [draftSystem, ...normalizedMessages];\n\n// Return a single item with fields used by downstream OpenAI HTTP Request nodes\nreturn [\n {\n json: {\n openaiModel,\n messages: normalizedMessages,\n userQuestion,\n messagesClarify,\n messagesDraft,\n },\n },\n];\n"
},
"typeVersion": 2
},
{
"id": "bd546b30-10b0-4b16-91c0-0eae052510b9",
"name": "OpenAI - Clarify",
"type": "n8n-nodes-base.httpRequest",
"position": [
608,
368
],
"parameters": {
"url": "https://api.openai.com/v1/chat/completions",
"method": "POST",
"options": {},
"jsonBody": "={{ {\n \"model\": $json.openaiModel,\n \"messages\": $json.messagesClarify,\n \"temperature\": 0\n} }}",
"sendBody": true,
"specifyBody": "json",
"authentication": "predefinedCredentialType",
"nodeCredentialType": "openAiApi"
},
"typeVersion": 4.3
},
{
"id": "2ee71615-4cb2-44b9-8930-872ec49adb37",
"name": "IF - Needs clarification?",
"type": "n8n-nodes-base.if",
"position": [
832,
368
],
"parameters": {
"options": {},
"conditions": {
"options": {
"version": 3,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"id": "155eb509-212b-44fb-8240-e6cb2c13c348",
"operator": {
"type": "string",
"operation": "notEquals"
},
"leftValue": "={{ $json.choices[0].message.content.trim() }}",
"rightValue": "END"
}
]
}
},
"typeVersion": 2.3
},
{
"id": "ef4c9f2e-6194-4994-8893-8a50d649a328",
"name": "Respond: Clarification Question",
"type": "n8n-nodes-base.respondToWebhook",
"position": [
1056,
272
],
"parameters": {
"options": {},
"respondWith": "json",
"responseBody": "={{ {\n status: 'needs_clarification',\n clarificationQuestion: $json.choices[0].message.content.trim()\n} }}"
},
"typeVersion": 1.5
},
{
"id": "0e62e22c-5af6-47c4-816b-147172e3f99d",
"name": "Open AI - Draft Answer",
"type": "n8n-nodes-base.httpRequest",
"position": [
992,
624
],
"parameters": {
"url": "https://api.openai.com/v1/chat/completions",
"method": "POST",
"options": {},
"jsonBody": "={{ {\n \"model\": $node['Prepare Input'].json.openaiModel,\n \"messages\": $node['Prepare Input'].json.messagesDraft,\n \"temperature\": 0.2\n} }}",
"sendBody": true,
"specifyBody": "json",
"authentication": "predefinedCredentialType",
"nodeCredentialType": "openAiApi"
},
"typeVersion": 4.3
},
{
"id": "50264f0c-d972-48ac-9661-3ddd01a958ca",
"name": "MCP Client",
"type": "@n8n/n8n-nodes-langchain.mcpClient",
"position": [
1216,
624
],
"parameters": {
"tool": {
"__rl": true,
"mode": "list",
"value": "verifyAnswer",
"cachedResultName": "verifyAnswer"
},
"options": {},
"inputMode": "json",
"jsonInput": "={{ {\n \"answer\": $json.choices[0].message.content.trim(),\n \"chatHistory\": $('Prepare Input').item.json.messages\n} }}",
"endpointUrl": "https://mcp.pearl.com/mcp",
"authentication": "headerAuth"
},
"typeVersion": 1
},
{
"id": "c5e89c86-8dbd-49ee-83ba-47b7c8f7454d",
"name": "Respond: Verified Answer by Pearl Expert",
"type": "n8n-nodes-base.respondToWebhook",
"position": [
1408,
624
],
"parameters": {
"options": {},
"respondWith": "json",
"responseBody": "={{ { \n status: 'verified', \n answer: $json.content[0].text.answer, \n verifiedBy: [\n $json.content?.[0]?.text?.expert?.name,\n $json.content?.[0]?.text?.expert?.expertise\n].filter(Boolean).join(', ')\n} }}"
},
"typeVersion": 1.5
},
{
"id": "76ac001e-e195-41af-89e8-8ead030182be",
"name": "Sticky Note",
"type": "n8n-nodes-base.stickyNote",
"position": [
96,
256
],
"parameters": {
"color": 7,
"width": 432,
"height": 272,
"content": "## 1) Intake & prompt setup\nReceives the user conversation and prepares prompts for intake + drafting.\n"
},
"typeVersion": 1
},
{
"id": "6e1aa795-0ac0-466c-904a-fa5272aa8db2",
"name": "Sticky Note1",
"type": "n8n-nodes-base.stickyNote",
"position": [
544,
256
],
"parameters": {
"color": 7,
"width": 704,
"height": 272,
"content": "## 2) Intake (clarification)\nAsks one follow-up question if more information is needed.\n"
},
"typeVersion": 1
},
{
"id": "2fb6d5ed-147a-4c96-8753-0c7f483aef4b",
"name": "Sticky Note2",
"type": "n8n-nodes-base.stickyNote",
"position": [
896,
544
],
"parameters": {
"color": 7,
"width": 704,
"height": 240,
"content": "## 3) Draft + expert verification\nDrafts an answer, then sends it to Pearl for professional verification.\n"
},
"typeVersion": 1
}
],
"active": false,
"settings": {
"availableInMCP": false,
"executionOrder": "v1"
},
"versionId": "3e782757-99b5-4cee-bd04-9eaaddad1869",
"connections": {
"MCP Client": {
"main": [
[
{
"node": "Respond: Verified Answer by Pearl Expert",
"type": "main",
"index": 0
}
]
]
},
"Prepare Input": {
"main": [
[
{
"node": "OpenAI - Clarify",
"type": "main",
"index": 0
}
]
]
},
"OpenAI - Clarify": {
"main": [
[
{
"node": "IF - Needs clarification?",
"type": "main",
"index": 0
}
]
]
},
"Pearl Verify Webhook": {
"main": [
[
{
"node": "Prepare Input",
"type": "main",
"index": 0
}
]
]
},
"Open AI - Draft Answer": {
"main": [
[
{
"node": "MCP Client",
"type": "main",
"index": 0
}
]
]
},
"IF - Needs clarification?": {
"main": [
[
{
"node": "Respond: Clarification Question",
"type": "main",
"index": 0
}
],
[
{
"node": "Open AI - Draft Answer",
"type": "main",
"index": 0
}
]
]
}
}
}
Pro
For the full experience including quality scoring and batch install features for each workflow upgrade to Pro
About this workflow
Deliver higher-confidence answers by combining an AI assistant with professional human verification.
Source: https://n8n.io/workflows/12725/ — original creator credit. Request a take-down →
Related workflows
Workflows that share integrations, category, or trigger type with this one. All free to copy and import.
Fully automated meeting documentation workflow that uses AI to transform raw transcripts into professional PDFs and actionable tasks. AI-powered summary generation (GPT-4) Automatic action item extrac
OpenAI, N8N Nodes Htmlcsstopdf, Gmail +4
This comprehensive workflow automates the entire testimonial collection and publishing process, from submission to social media-ready content. It uses AI to enhance testimonials, generates beautiful b
OpenAI, N8N Nodes Htmlcsstoimage, HTTP Request +3
Automatically transform resolved support tickets into professional, AI-powered PDF documentation with complete tracking and team notifications.
OpenAI, Google Drive, Google Sheets +3
Automated SEO Audit in n8n – Your All-in-One Website Optimization Tool!
HTTP Request, Html Extract, Email Send +2
Simplify employee performance reviews with AI-powered automation. This workflow transforms raw feedback and evaluation inputs into clear, structured, and professional performance review summaries — sa
OpenAI, Gmail, HTTP Request +3