> For the complete documentation index, see [llms.txt](https://docs.clickai.vn/clickai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.clickai.vn/clickai-docs/clickai-docs-en/function/database/integration-with-ai.md). # Integration with AI ## Table of Contents · \[Integration Overview]\(#integration-overview) · \[Tables + Workflow]\(#tables--workflow) · \[Tables + AI Agent]\(#tables--ai-agent) · \[Tables + Webhooks]\(#tables--webhooks) · \[REST API]\(#rest-api) · \[MCP Server — Connect AI via MCP Protocol]\(#mcp-server--connect-ai-via-mcp-protocol) · \[AI Integration Use-cases]\(#ai-integration-use-cases) · \[Best Practices]\(#best-practices) ## Integration Overview ClickAI Tables is more than just data storage — it's the central hub connecting the entire ClickAI AI ecosystem. You can: graph LR\ A\[📊 Tables] --> B\[🤖 AI Agent]\ A --> C\[⚙️ Workflow]\ A --> D\[🔔 Webhooks]\ A --> E\[📡 REST API]\ A --> F\[🔌 MCP Server]\ C --> A\ B --> A\ F --> B · Read data from Tables in Workflows/Agents · Write data to Tables from AI processing results · Trigger automations when data changes via Webhooks · Access programmatically via REST API · Connect LLMs directly via MCP Server (Model Context Protocol) ## Tables + Workflow ### Read Data from Tables in a Workflow Use the HTTP Request node in a Workflow to query data from the Tables API: 1\. Add an HTTP Request node to the Workflow 2\. Configure: · Method: GET · URL: · Headers: xc-auth: {API\_TOKEN} 3\. Process the response data in subsequent nodes ### Write Data to Tables from a Workflow 4\. Add an HTTP Request node to the Workflow 5\. Configure: · Method: POST · URL: · Headers: xc-auth: {API\_TOKEN} · Body: JSON containing the new record data {\ "Title": "{{workflow\.output.title}}",\ "Status": "Processed",\ "AI\_Summary": "{{workflow\.output.summary}}",\ "Score": {{workflow\.output.score}}\ } ### Example Workflow: Automated Lead Processing graph TD\ A\[🔔 Webhook: New Lead] --> B\[📊 Read Lead from Tables]\ B --> C\[🧠 LLM: Analyze Lead]\ C --> D{Score > 80?}\ D -->|Yes| E\[📊 Update Status = Hot]\ D -->|No| F\[📊 Update Status = Cold]\ E --> G\[📧 Send notification to Sales]\ F --> H\[📧 Send nurturing email] ## Tables + AI Agent ### Use Tables as a Tool for AI Agent AI Agents can use the Tables API as a tool to query and update data: 6\. In the Agent configuration, add a Custom Tool 7\. Configure the Tables API endpoint 8\. The Agent will automatically call the API when it needs to look up information Example: A customer support Agent can: · Look up customer information from the "Customers" table · Check order status from the "Orders" table · Create new tickets in the "Support Tickets" table · Update processing status ### Connect Tables via App Integration Use the App Integration module (Composio) to connect Tables: 9\. Go to Workspace → App Integration 10\. Find and connect NocoDB/ClickAI Tables 11\. Configure authentication 12\. Use the available actions in Agent/Workflow ## Tables + Webhooks Webhooks send automatic notifications when data in Tables changes: ### Webhook Configuration 13\. Open the table → Table Details → Webhooks tab 14\. Click Create Webhook 15\. Configure:


Setting	Description
Event	After Insert / After Update / After Delete
URL	Receiving endpoint (e.g., ClickAI Workflow trigger)
Method	POST
Headers	Custom headers if needed
Condition	Trigger only when specific conditions are met

### Webhook Payload {\ "type": "records.after.insert",\ "data": {\ "table\_id": "tbl\_xxxxx",\ "rows": \[\ {\ "Id": 1,\ "Title": "New Customer Lead",\ "Email": "",\ "Status": "New"\ }\ ]\ },\ "timestamp": "2024-01-15T10:30:00Z"\ } ### Connect Webhook → ClickAI Workflow 16\. Create a new Workflow in ClickAI Studio 17\. Use the Start Node with trigger type Webhook 18\. Copy the Webhook URL from the Workflow 19\. Paste the URL into the table's Webhook configuration 20\. Whenever data changes → Webhook triggers → Workflow runs automatically ## REST API ClickAI Tables provides a full REST API for all data operations: ### Authentication \# Using API Token\ curl -H "xc-auth: YOUR\_API\_TOKEN" \\\ ### Key API Endpoints


Method	Endpoint	Description
GET	/api/v2/tables/{id}/records	List records
GET	/api/v2/tables/{id}/records/{row_id}	Get a specific record
POST	/api/v2/tables/{id}/records	Create new record(s)
PATCH	/api/v2/tables/{id}/records	Update record(s)
DELETE	/api/v2/tables/{id}/records	Delete record(s)

### Query Parameters


Parameter	Description	Example
where	Filter records	(Status,eq,Active)
sort	Sort records	-CreatedAt (descending)
fields	Select returned fields	Title,Email,Status
limit	Limit number of records	25
offset	Pagination offset	50

### Example: Get Active Customers curl -X GET \\\ "" \\\ -H "xc-auth: YOUR\_API\_TOKEN" ### Example: Create a New Record curl -X POST \\\ "" \\\ -H "xc-auth: YOUR\_API\_TOKEN" \\\ -H "Content-Type: application/json" \\\ -d '{\ "Title": "John Doe",\ "Email": "",\ "Phone": "+1234567890",\ "Status": "New"\ }' 💡 TIP: You can find the API Token in \*\*Table Details\*\* → \*\*API Snippet\*\* tab or in \*\*Workspace Settings\*\* → \*\*API Tokens\*\*. ## MCP Server — Connect AI via MCP Protocol MCP (Model Context Protocol) is a standard protocol that allows AI applications (LLMs) to interact directly with data in Tables using natural language — no API coding required. With MCP, you can ask AI directly: · \*"Show me all customers from Hanoi"\* · \*"Add a new task: Review documentation, priority High"\* · \*"Update order #123 status to Delivered"\* · \*"Delete all leads with status = Lost"\* ### How MCP Works graph LR\ A\[👤 User] -->|Natural language| B\[🤖 AI Client]\ B -->|MCP Protocol| C\[🔌 MCP Server]\ C -->|REST API| D\[📊 ClickAI Tables]\ D -->|Data| C\ C -->|Results| B\ B -->|Response| A


Component	Role
AI Client	AI application (Claude Desktop, Cursor, Windsurf...)
MCP Server	Bridge between AI and Tables — translates AI commands into API calls
ClickAI Tables	Database that receives and processes data requests

### MCP Server Capabilities


Feature Group	Capabilities
Record CRUD	Create, read, update, delete records using natural language
Table Management	List, create, delete tables
Field Management	Add, delete, manage fields/columns
View Management	List, create, delete views (Grid, Gallery, Kanban...)
Data Operations	Filter, sort, search, aggregate data

### Configure MCP on ClickAI Tables #### Method 1: Create MCP Endpoint from the UI (Recommended) 21\. Open a base in ClickAI Tables 22\. Click the Overview button in the left sidebar 23\. Select the Settings tab 24\. Choose Model Context Protocol from the settings menu 25\. Click New MCP Endpoint 26\. Name the endpoint and click Create 27\. Copy the auto-generated MCP Config JSON #### Method 2: Manual Configuration (Self-hosted) For self-hosted ClickAI Tables instances, configure manually using an API Token: Step 1: Gather configuration details


Setting	How to obtain
NOCODB_BASE_URL	Your Tables instance URL (e.g., https://tables.clickai.vn)
NOCODB_API_TOKEN	Account Settings → API Tokens → Create Token
NOCODB_BASE_ID	Found in the URL when opening a base: .../#/nc/base/BASE_ID_HERE/...

Step 2: Create the MCP Config JSON {\ "mcpServers": {\ "clickai-tables": {\ "command": "npx",\ "args": \["-y", "nocodb-mcp-server"],\ "env": {\ "NOCODB\_BASE\_URL": " "NOCODB\_API\_TOKEN": "your\_api\_token\_here",\ "NOCODB\_BASE\_ID": "your\_base\_id\_here"\ }\ }\ }\ } ### Connect to Claude Desktop 28\. Open Claude Desktop → Preferences (⌘+, on Mac / Ctrl+, on Windows) 29\. Go to Develop → Edit Config 30\. Paste the MCP Config JSON into claude\_desktop\_config.json 31\. Save the file and restart Claude Desktop 32\. Verify: open Claude, you should see the 🔌 MCP Tools icon in the input area // \~/Library/Application Support/Claude/claude\_desktop\_config.json (macOS)\ // %APPDATA%\Claude\claude\_desktop\_config.json (Windows)\ {\ "mcpServers": {\ "clickai-tables": {\ "command": "npx",\ "args": \["-y", "nocodb-mcp-server"],\ "env": {\ "NOCODB\_BASE\_URL": " "NOCODB\_API\_TOKEN": "xc-auth-xxxxxxxxxxxxx",\ "NOCODB\_BASE\_ID": "p\_xxxxxxxxxxxxxxxx"\ }\ }\ }\ } ### Connect to Cursor 33\. Open Cursor → Settings (⇧+⌘+J / Ctrl+Shift+J) 34\. Navigate to the MCP tab 35\. Click Add Custom MCP 36\. Paste the MCP Config JSON and save 37\. On success, you will see the NocoDB tools listed ### Example Usage with Claude After connecting, you can communicate using natural language: 👤 "Show me the 10 most recent customers in the Customers table"\ 🤖 Claude auto-calls MCP → list\_records(table: Customers, sort: -CreatedAt, limit: 10)\ → Returns a table of the 10 newest customers\ \ 👤 "Add a new customer: John Doe, email: , phone: +1234567890"\ 🤖 Claude auto-calls MCP → create\_record(table: Customers, data: {...})\ → ✅ Record created successfully\ \ 👤 "Create a new table called 'Feedback' with 3 columns: Customer Name, Rating, Comment"\ 🤖 Claude auto-calls MCP → create\_table(...) + create\_fields(...)\ → ✅ Feedback table created with 3 fields ### MCP Permission Management


Mode	Description	Best for
Always Ask Permission	AI asks for confirmation before each action	Production data, critical operations
Allow Unsupervised	AI executes without confirmation	Test data, dev environments

🚨 WARNING: \*\*Security when using MCP:\*\* - Use API Tokens with minimal permissions (read-only if you only need to query) - Never use Admin tokens for MCP integrations - Enable "Always Ask Permission" mode for production data - Review MCP Config JSON carefully before sharing — ensure tokens are not exposed 💡 TIP: MCP is especially useful for developers who want to quickly prototype, query data, or manage databases without opening the Tables UI. Combining MCP with Claude Desktop turns Tables into your personal data assistant. ## AI Integration Use-cases ### 1. Smart CRM with AI Agent Customer chats → AI Agent queries CRM in Tables →\ Responds based on purchase history → Logs interaction in Tables ### 2. Automated Customer Segmentation Webhook: New Order → Workflow → LLM analyzes behavior →\ Updates Segment field in Tables → Triggers personalized outreach ### 3. AI-Powered Content Calendar Form View collects ideas → Workflow → LLM writes draft →\ Saves draft to table → Kanban View manages progress ### 4. Automated Helpdesk Form View receives ticket → Webhook → AI classifies →\ Updates Priority & Category → Assigns to appropriate staff ### 5. Intelligent Reporting Scheduled Workflow → Query Tables API → LLM analyzes data →\ Generates report → Sends email to management ## Best Practices 💡 TIP: \*\*Effective Table Design:\*\* - Each table should focus on a single entity (Customers, Orders, Products) - Use Links to create relationships instead of duplicating data - Use clear, descriptive field names - Use Single Select for status fields to leverage Kanban View 🚨 WARNING: \*\*API Integration Notes:\*\* - Secure API Tokens — never hardcode in client-side code - Implement rate limiting to avoid excessive API calls - Validate data before writing to Tables - Set up error handling in Workflows for API failure cases ⚠️ IMPORTANT: \*\*Performance Optimization:\*\* - Use the \`fields\` parameter to fetch only necessary data - Implement pagination for tables with many records - Use \`where\` filters at the API level instead of client-side filtering - Cache API results when data doesn't change frequently --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://docs.clickai.vn/clickai-docs/clickai-docs-en/function/database/integration-with-ai.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.