This workflow follows the Agent → Chat Trigger recipe pattern — see all workflows that pair these two integrations.
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 →
{
"nodes": [
{
"id": "53b36910-966f-45ba-a425-a3260a55059f",
"name": "OpenAI Chat Model",
"type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
"position": [
340,
480
],
"parameters": {
"model": {
"__rl": true,
"mode": "list",
"value": "gpt-4o-mini"
},
"options": {}
},
"typeVersion": 1.2
},
{
"id": "177235e8-c925-43d0-9695-10f072e26350",
"name": "AI Control Tower Agent",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
380,
240
],
"parameters": {
"options": {
"systemMessage": "=You are an AI-powered SQL assistant specialized in supply chain analytics. \nYour role is to execute SQL queries on BigQuery and return only the results in a structured format.\n\nToday we are May 31, 2021.\n\n### **Behavior & Rules**\n1\ufe0f\u20e3 **Query Execution:**\n - Your only task is to process user requests and return **direct results** from BigQuery.\n - Do **not** display the SQL query.\n - Only return structured **data** as output.\n\n2\ufe0f\u20e3 **Data Presentation:**\n - Format the results as a **table** whenever possible.\n - If results are numerical (counts, percentages, aggregates), return them **clearly and concisely**.\n - If results contain multiple rows, return **only the first 10** for preview, unless the user specifies otherwise.\n\n3\ufe0f\u20e3 **Handling Large Datasets:**\n - If the user asks for many rows, show the first **100 rows max** unless specified.\n - Provide a **summary** when dealing with large data instead of showing everything.\n\n4\ufe0f\u20e3 **Response Format:**\n - \u2705 **For counts & metrics:** \n `\"There were 5,432 delayed shipments in the last 21 days.\"`\n - \u2705 **For tables:** \n | ShipmentID | City | Store | Order Date | Delivery Date | On Time? |\n |-----------|-------|--------|------------|--------------|----------|\n | 12345 | NYC | ST1 | 2024-03-10 | 2024-03-15 | No |\n | 67890 | Paris | ST4 | 2024-03-11 | 2024-03-16 | Yes |\n\n5\ufe0f\u20e3 **Clarifying Unclear Requests:**\n - If the user request is **too broad**, ask for clarification instead of running an expensive query.\n\n---\n\n### Schema Awareness\nAll SQL queries must use the BigQuery table: \n`transport.shipments` \n\nThis table includes fields such as:\n- `Shipment ID`, `City`, `Store`, `Order Date`, `Delivery Date`, `On Time Delivery`\n- As well as operational timestamps: `Transmission`, `Loading`, `Airport Arrival`, etc.\n- And status flags: `Transmission OnTime`, `Loading OnTime`, `Airport OnTime`, `Store Open`\n\nUse these fields appropriately when analyzing shipment performance.\n\n---\n\n### Tool Usage Instruction (for \"bigquery_tool\")\n\nWhenever you need to run a SQL query, use the tool called `bigquery_tool`.\n\nYou must provide the query in the following format:\n```json\n{\n \"query\": \"SELECT COUNT(*) FROM `transport.shipments` WHERE `On Time Delivery` = FALSE\"\n}\n"
}
},
"typeVersion": 1.8
},
{
"id": "5366cc5f-85d3-44d2-9b1b-62febfcb44e3",
"name": "Sticky Note1",
"type": "n8n-nodes-base.stickyNote",
"position": [
-100,
-120
],
"parameters": {
"color": 7,
"width": 200,
"height": 520,
"content": "### 1. Workflow Trigger with Chat\nThis workflow uses a simple chat window as a trigger. You can replace it with Telegram, Slack, Teams or a webhook trigger linked to your chat.\n\n#### How to setup?\n*Nothing to do.*\n"
},
"typeVersion": 1
},
{
"id": "4218a062-12f8-437d-ab22-5a653a3089b2",
"name": "Sticky Note2",
"type": "n8n-nodes-base.stickyNote",
"position": [
140,
-120
],
"parameters": {
"color": 7,
"width": 700,
"height": 740,
"content": "### 2. AI Agent equipped with the query tool\nIn order to have more control on the input of the BigQuery node, we don't use the BigQuery tool. Instead we have a set of nodes to retrieve the SQL query, clean it and send it to a BigQuery Node.\n\n#### How to setup?\n- **AI Agent with the Chat Model**:\n 1. Add a **chat model** with the required credentials *(Example: Open AI 4o-mini)*\n 2. Adapt the **name of your BigQuery table** in the system prompt *(Example: transports.shipments)*\n 3. Adapt the **tables fields explanation** in the system prompt\n [Learn more about the AI Agent Node](https://docs.n8n.io/integrations/builtin/cluster-nodes/root-nodes/n8n-nodes-langchain.agent)\n- Copy and past the **nodes in the yellow sticker** in another workflow. Point the query tool to this workflow.\n[Learn more about the Custom n8n Workflow Tool node](https://docs.n8n.io/integrations/builtin/cluster-nodes/sub-nodes/n8n-nodes-langchain.toolworkflow)"
},
"typeVersion": 1
},
{
"id": "c5967f58-00e8-4f03-9110-913547f7ab9c",
"name": "Call Query Tool",
"type": "@n8n/n8n-nodes-langchain.toolWorkflow",
"position": [
640,
440
],
"parameters": {
"name": "bigquery_tool",
"workflowId": {
"__rl": true,
"mode": "list",
"value": "4Os7DoxHjFuTwWio",
"cachedResultName": "\ud83d\udd28 Big Query Tool"
},
"description": "=Use this tool to run an SQL query and fetch the result from the BigQuery database.\n\nThe tool expects input in the following format:\n{\n \"query\": \"SELECT COUNT(*) FROM `transport.shipments` WHERE `On Time Delivery` = FALSE\"\n}\n\nOnly provide the SQL query as a string inside the 'query' key. Do not include code formatting (like ```sql), comments, or explanations. The tool will return only the raw result from the database.\n",
"workflowInputs": {
"value": {
"query": "={{ $fromAI(\"query\", \"SQL query to run\") }}"
},
"schema": [
{
"id": "query",
"type": "string",
"display": true,
"removed": false,
"required": false,
"displayName": "query",
"defaultMatch": false,
"canBeUsedToMatch": true
}
],
"mappingMode": "defineBelow",
"matchingColumns": [
"query"
],
"attemptToConvertTypes": false,
"convertFieldsToString": false
}
},
"typeVersion": 2
},
{
"id": "429813c8-b07f-4551-aeea-1744a1225449",
"name": "Sticky Note",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
-120
],
"parameters": {
"width": 760,
"height": 460,
"content": "### 3. Big Query Workflow\nExecute the SQL query generated by the AI agent in Big Query. Retrieve the results and send them back to the AI Agent.\n\n### How to set up?\n- Paste these nodes in a separate workflow so you can use it with multiple agents.\n- **Google BigQuery API**:\n 1. Add your Google Translate API credentials\n 2. The project in which your table is located\n [Learn more about the Google BigQuery Node](https://docs.n8n.io/integrations/builtin/app-nodes/n8n-nodes-base.googlebigquery)\n"
},
"typeVersion": 1
},
{
"id": "bede0624-8923-4af0-8adc-8be22d556066",
"name": "Query Database",
"type": "n8n-nodes-base.googleBigQuery",
"position": [
1520,
180
],
"parameters": {
"options": {},
"sqlQuery": "={{ $json.query }}",
"projectId": {
"__rl": true,
"mode": "list",
"value": "=",
"cachedResultUrl": "=",
"cachedResultName": "="
}
},
"notesInFlow": true,
"typeVersion": 2.1
},
{
"id": "137e4dbc-db8d-4ec7-a3e0-478dde6ef27c",
"name": "Trigger Executed by the AI Tool",
"type": "n8n-nodes-base.executeWorkflowTrigger",
"position": [
960,
180
],
"parameters": {
"workflowInputs": {
"values": [
{
"name": "query"
}
]
}
},
"typeVersion": 1.1
},
{
"id": "42a2801e-582e-4340-83af-ef0041eab4f9",
"name": "Sanitising the Query",
"type": "n8n-nodes-base.code",
"position": [
1240,
180
],
"parameters": {
"jsCode": "return [\n {\n json: {\n query: $input.first().json.query.replace(/```sql|```/g, \"\").trim()\n }\n }\n];\n"
},
"typeVersion": 2
},
{
"id": "7c86fda0-116c-47ad-aaf5-8b83d2c083c6",
"name": "Chat Memory",
"type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
"position": [
480,
480
],
"parameters": {},
"typeVersion": 1.3
},
{
"id": "e1408ac1-24da-4d38-8fdf-c110a54d3f55",
"name": "Chat with the User",
"type": "@n8n/n8n-nodes-langchain.chatTrigger",
"position": [
-60,
240
],
"parameters": {
"options": {}
},
"typeVersion": 1.1
},
{
"id": "bc49829b-45f2-4910-9c37-907271982f14",
"name": "Sticky Note3",
"type": "n8n-nodes-base.stickyNote",
"position": [
900,
380
],
"parameters": {
"width": 780,
"height": 540,
"content": "### 4. Do you need more details?\nFind a step-by-step guide in this tutorial\n\n[\ud83c\udfa5 Watch My Tutorial](https://www.loom.com/share/50271f9d50214d7184830985497a75ec?sid=d0c410dc-29f1-488f-b89a-4011de0ded07)"
},
"typeVersion": 1
}
],
"connections": {
"Chat Memory": {
"ai_memory": [
[
{
"node": "AI Control Tower Agent",
"type": "ai_memory",
"index": 0
}
]
]
},
"Call Query Tool": {
"ai_tool": [
[
{
"node": "AI Control Tower Agent",
"type": "ai_tool",
"index": 0
}
]
]
},
"OpenAI Chat Model": {
"ai_languageModel": [
[
{
"node": "AI Control Tower Agent",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"Chat with the User": {
"main": [
[
{
"node": "AI Control Tower Agent",
"type": "main",
"index": 0
}
]
]
},
"Sanitising the Query": {
"main": [
[
{
"node": "Query Database",
"type": "main",
"index": 0
}
]
]
},
"Trigger Executed by the AI Tool": {
"main": [
[
{
"node": "Sanitising the Query",
"type": "main",
"index": 0
}
]
]
}
}
}
Pro
For the full experience including quality scoring and batch install features for each workflow upgrade to Pro
How this works
This workflow empowers users to query Google BigQuery databases through natural language conversations, delivering precise data insights without needing SQL expertise. It suits data analysts, business intelligence teams, or anyone managing large datasets who want quick, AI-assisted responses to complex questions. The core step involves an AI agent interpreting your query via OpenAI's chat model, then executing it against BigQuery to return structured results, all while using sticky notes to maintain context across interactions.
Use this workflow for ad-hoc data exploration in event-driven scenarios, such as real-time dashboard updates or on-demand reporting, especially when integrating with tools like memory buffers for session persistence. Avoid it for high-volume batch processing or scenarios requiring strict SQL customisation, where direct scripting might be more efficient. Common variations include adding workflow triggers for scheduled queries or chaining multiple agents for multi-step analytics.
About this workflow
Googlebigquery Stickynote. Uses lmChatOpenAi, agent, stickyNote, toolWorkflow. Event-driven trigger; 12 nodes.
Source: https://github.com/Zie619/n8n-workflows — 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.
Splitout Redis. Uses executeWorkflowTrigger, n8n, redis, splitOut. Event-driven trigger; 46 nodes.
Execute Workflow Trigger, n8n, Redis +7
3770. Uses executeWorkflowTrigger, n8n, redis, agent. Event-driven trigger; 46 nodes.
Execute Workflow Trigger, n8n, Redis +7
Designing agent tools for outcome rather than utility has been a long recommended practice of mine and it applies well when it comes to building MCP servers; In gist, agents to be making the least amo
Execute Workflow Trigger, n8n, Redis +7
Workflow 3770. Uses executeWorkflowTrigger, n8n, redis, agent. Event-driven trigger; 46 nodes.
Execute Workflow Trigger, n8n, Redis +7
Build your own N8N Workflows MCP Server. Uses executeWorkflowTrigger, n8n, redis, agent. Event-driven trigger; 46 nodes.
Execute Workflow Trigger, n8n, Redis +7