dingtalk-openclaw-connector

name	dingtalk-openclaw-connector
description	Connect OpenClaw AI agents to DingTalk with message handling, document operations, calendar, todos, and AI cards
triggers	["how do I integrate OpenClaw with DingTalk","set up DingTalk connector for OpenClaw","create a DingTalk bot with OpenClaw","how to handle DingTalk messages in OpenClaw","send DingTalk notifications from OpenClaw agent","configure DingTalk AI card streaming","troubleshoot DingTalk OpenClaw connector","route multiple DingTalk bots to different agents"]

DingTalk OpenClaw Connector Skill

Skill by ara.so — Hermes Skills collection.

Official DingTalk channel plugin for OpenClaw, enabling AI agents to receive/send messages, manage documents, calendars, todos, and more within DingTalk enterprise workspace.

What It Does

This TypeScript connector bridges OpenClaw agents with DingTalk's ecosystem:

• Messaging: Receive/send private/group messages, @mentions, rich media (text/Markdown/images)
• Documents: Create, append, search, list DingTalk documents
• DING Notifications: Send priority alerts to users/groups
• Todos: Create personal todos, check status, set deadlines
• AI Tables: Create tables, read/write rows, query data
• Calendar: Manage calendars, events (CRUD + search), attendees, availability
• Journal: Submit daily/weekly reports, query history
• AI Cards: Streaming responses with typewriter effect, interactive buttons
• Multi-Agent Routing: Connect multiple bots to different OpenClaw agents
• Access Control: Flexible permission policies for private/group chats

Prerequisites

• OpenClaw ≥ 2026.4.9 (check: openclaw -v, upgrade: npm install -g openclaw)
• Node.js (for installation)
• DingTalk App (mobile, for QR authorization)

Installation

Quick Install with Auto-Authorization

npx -y @dingtalk-real-ai/dingtalk-connector install

Scan the QR code displayed in terminal with DingTalk mobile app. After seeing "Success! Bot configured.", restart gateway:

openclaw gateway restart

Manual Installation (if auto-auth fails)

npm install -g @dingtalk-real-ai/dingtalk-connector

Then follow manual setup guide to configure credentials.

Configuration

Plugin config is stored in OpenClaw's channel settings. Key environment variables:

# DingTalk bot credentials (obtained during authorization)
DINGTALK_CLIENT_ID=your_client_id
DINGTALK_CLIENT_SECRET=your_client_secret
DINGTALK_ROBOT_CODE=your_robot_code

# Optional: Multi-agent routing
DINGTALK_AGENT_MAPPING='{"bot_code_1":"agent_1","bot_code_2":"agent_2"}'

# Optional: Access control
DINGTALK_PRIVATE_CHAT_POLICY=whitelist  # whitelist|blacklist|all
DINGTALK_GROUP_CHAT_POLICY=all          # whitelist|blacklist|all
DINGTALK_WHITELIST=user_id_1,user_id_2

Access Control Policies

Configure who can interact with your bot:

// In OpenClaw channel config
{
  "privateChatPolicy": "whitelist",  // Only whitelisted users
  "groupChatPolicy": "all",          // All groups
  "whitelist": ["user_123", "user_456"],
  "blacklist": []
}

Options: all, whitelist, blacklist

Key Capabilities

1. Message Handling

The connector automatically receives and routes DingTalk messages to your OpenClaw agent:

// Agent receives message context
interface MessageContext {
  conversationId: string;
  senderId: string;
  senderName: string;
  content: {
    text?: string;
    images?: Array<{ downloadCode: string; url: string }>;
    files?: Array<{ fileName: string; downloadCode: string }>;
  };
  isGroupChat: boolean;
  atUsers?: string[];
}

2. Sending Messages

Text Message:

// Agent action
{
  "action": "sendMessage",
  "params": {
    "conversationId": "cid_xxx",
    "content": "Hello from OpenClaw!"
  }
}

Markdown with @mention:

{
  "action": "sendMessage",
  "params": {
    "conversationId": "cid_xxx",
    "content": "## Report\n@user_123 please review",
    "format": "markdown",
    "atUsers": ["user_123"]
  }
}

Image Message:

{
  "action": "sendMessage",
  "params": {
    "conversationId": "cid_xxx",
    "imageUrl": "https://example.com/image.png"
    // Or local path: "imagePath": "/path/to/image.png"
  }
}

3. AI Card Streaming

Show real-time thinking/generation progress:

// Enable in channel config
{
  "enableAICard": true,
  "streamingMode": "typewriter"  // typewriter effect
}

The connector automatically wraps agent responses in interactive cards with states:

• 🤔 Thinking...
• ✍️ Generating...
• ✅ Complete

4. Document Operations

Create Document:

{
  "action": "createDocument",
  "params": {
    "title": "Meeting Notes",
    "content": "# Agenda\n- Item 1\n- Item 2",
    "spaceId": "space_xxx"  // optional
  }
}

Append to Document:

{
  "action": "appendDocument",
  "params": {
    "documentId": "doc_xxx",
    "content": "\n## New Section\nAdditional notes..."
  }
}

Search Documents:

{
  "action": "searchDocuments",
  "params": {
    "keyword": "meeting",
    "maxResults": 10
  }
}

5. DING Notifications

Send high-priority alerts:

{
  "action": "sendDing",
  "params": {
    "receiverUserIds": ["user_123", "user_456"],
    "content": "Urgent: Server down!",
    "remindType": "DING_SMS"  // DING_NOTICE or DING_SMS
  }
}

6. Todo Management

Create Todo:

{
  "action": "createTodo",
  "params": {
    "subject": "Review PR #123",
    "description": "Check code quality",
    "dueTime": "2026-05-20T17:00:00Z",
    "executorIds": ["user_123"]
  }
}

Query Todos:

{
  "action": "getTodos",
  "params": {
    "status": "PENDING",  // PENDING|DONE
    "startDate": "2026-05-01",
    "endDate": "2026-05-31"
  }
}

7. Calendar & Events

Create Calendar Event:

{
  "action": "createCalendarEvent",
  "params": {
    "calendarId": "cal_xxx",
    "summary": "Team Sync",
    "startTime": "2026-05-20T14:00:00+08:00",
    "endTime": "2026-05-20T15:00:00+08:00",
    "location": "Conference Room A",
    "attendees": [
      { "userId": "user_123" },
      { "userId": "user_456" }
    ]
  }
}

Query Events:

{
  "action": "searchCalendarEvents",
  "params": {
    "calendarId": "cal_xxx",
    "startTime": "2026-05-20T00:00:00+08:00",
    "endTime": "2026-05-21T00:00:00+08:00"
  }
}

Check Availability:

{
  "action": "checkAvailability",
  "params": {
    "userIds": ["user_123", "user_456"],
    "startTime": "2026-05-20T14:00:00+08:00",
    "endTime": "2026-05-20T15:00:00+08:00"
  }
}

8. AI Tables

Create Table:

{
  "action": "createAITable",
  "params": {
    "name": "Customer Database",
    "fields": [
      { "name": "Name", "type": "TEXT" },
      { "name": "Email", "type": "TEXT" },
      { "name": "Status", "type": "SINGLE_SELECT", "options": ["Active", "Inactive"] }
    ]
  }
}

Insert Row:

{
  "action": "insertTableRow",
  "params": {
    "tableId": "tbl_xxx",
    "fields": {
      "Name": "John Doe",
      "Email": "[e-mail]",
      "Status": "Active"
    }
  }
}

Query Rows:

{
  "action": "queryTableRows",
  "params": {
    "tableId": "tbl_xxx",
    "filter": {
      "Status": "Active"
    },
    "maxResults": 50
  }
}

9. Journal (Reports)

Submit Daily Report:

{
  "action": "submitJournal",
  "params": {
    "type": "daily",
    "date": "2026-05-20",
    "content": "## Completed\n- Feature A\n- Bug fix B\n\n## Tomorrow\n- Feature C"
  }
}

Query Reports:

{
  "action": "getJournals",
  "params": {
    "type": "weekly",
    "startDate": "2026-05-01",
    "endDate": "2026-05-15"
  }
}

Multi-Agent Routing

Connect multiple DingTalk bots to different OpenClaw agents for specialized tasks:

Configuration

# In .env or OpenClaw config
DINGTALK_AGENT_MAPPING='{
  "robot_code_hr": "hr_agent",
  "robot_code_it": "it_support_agent",
  "robot_code_sales": "sales_agent"
}'

Agent Definitions

# openclaw.config.yaml
agents:
  hr_agent:
    name: HR Assistant
    model: gpt-4
    systemPrompt: You are an HR assistant handling employee queries
    
  it_support_agent:
    name: IT Support
    model: gpt-4
    systemPrompt: You provide IT technical support
    
  sales_agent:
    name: Sales Helper
    model: gpt-4
    systemPrompt: You assist with sales inquiries and CRM

Each bot routes to its designated agent automatically. See Multi-Agent Setup Guide.

Common Patterns

Pattern 1: Auto-Reply Bot

// Agent receives all group messages where bot is @mentioned
// Auto-respond with context-aware answers
async function handleMessage(context: MessageContext) {
  const { content, isGroupChat, atUsers } = context;
  
  if (isGroupChat && !atUsers?.includes(botUserId)) {
    return; // Ignore if not @mentioned
  }
  
  const response = await generateResponse(content.text);
  
  return {
    action: "sendMessage",
    params: {
      conversationId: context.conversationId,
      content: response
    }
  };
}

Pattern 2: Document Search Assistant

// Search docs and create summary
async function searchAndSummarize(query: string) {
  // 1. Search documents
  const docs = await executeAction({
    action: "searchDocuments",
    params: { keyword: query, maxResults: 5 }
  });
  
  // 2. Generate summary
  const summary = await summarize(docs);
  
  // 3. Create new doc with results
  return executeAction({
    action: "createDocument",
    params: {
      title: `Search Results: ${query}`,
      content: summary
    }
  });
}

Pattern 3: Meeting Scheduler

// Find available slot and create event
async function scheduleMeeting(attendeeIds: string[], duration: number) {
  const now = new Date();
  const nextWeek = new Date(now.getTime() + 7 * 24 * 60 * 60 * 1000);
  
  // 1. Check availability
  const availability = await executeAction({
    action: "checkAvailability",
    params: {
      userIds: attendeeIds,
      startTime: now.toISOString(),
      endTime: nextWeek.toISOString()
    }
  });
  
  // 2. Find first free slot
  const freeSlot = findFreeSlot(availability, duration);
  
  // 3. Create event
  return executeAction({
    action: "createCalendarEvent",
    params: {
      summary: "Team Meeting",
      startTime: freeSlot.start,
      endTime: freeSlot.end,
      attendees: attendeeIds.map( ({ : id }))
    }
  });
}

Pattern 4: Task Tracker Bot

// Create todo from message command
async function createTaskFromMessage(message: string, conversationId: string) {
  // Parse: "todo: Review PR #123 by Friday"
  const match = message.match(/todo:\s*(.+?)\s+by\s+(.+)/i);
  if (!match) return;
  
  const [, task, deadline] = match;
  
  // Create todo
  const todo = await executeAction({
    action: "createTodo",
    params: {
      subject: task,
      dueTime: parseDeadline(deadline)
    }
  });
  
  // Confirm in chat
  return executeAction({
    action: "sendMessage",
    params: {
      conversationId,
      content: `✅ Task created: ${task}\n📅 Due: ${deadline}`
    }
  });
}

Troubleshooting

Bot Not Responding

1. Check gateway status:

   openclaw gateway status
   

2. Restart gateway:

   openclaw gateway restart
   

3. Verify credentials:

   # Check if config exists
   openclaw config list
   

4. Check logs:

   openclaw gateway logs
   

Authorization Failed

• Ensure you're using DingTalk mobile app (not desktop) to scan QR
• Check if QR code expired (timeout ~5 min), restart installation
• Verify network access to DingTalk API endpoints

Message Not Received

• Confirm bot is in the conversation (group chat: must add bot; private: user initiated)
• Check access control policies (whitelist/blacklist)
• Verify bot has required permissions in DingTalk admin console

Streaming AI Card Not Working

// Ensure enabled in config
{
  "enableAICard": true,
  "streamingMode": "typewriter"
}

• Feature requires OpenClaw ≥ 2026.4.9
• Check browser console for WebSocket errors

Multi-Agent Routing Issues

• Verify DINGTALK_AGENT_MAPPING JSON syntax
• Ensure agent names match those in openclaw.config.yaml
• Check robot codes are correct (not client IDs)

Image Upload Fails

• Ensure image file exists and is readable
• Check file size (DingTalk limit: 20MB for images)
• Verify image format (JPEG, PNG, GIF supported)
• For URLs, ensure publicly accessible

Calendar Events Not Creating

• Verify user has calendar permissions in DingTalk
• Check time format (ISO 8601 with timezone)
• Ensure calendarId is valid (create calendar first if needed)

Common Error Codes

• 40014: Invalid access token → Re-authorize bot
• 60020: Permission denied → Check DingTalk app permissions
• 90002: Message frequency limit → Rate limiting, slow down
• 71006: Robot not in conversation → Add bot to group/chat

Testing

Quick test after installation:

# Send test message to bot in DingTalk
# Expected: Bot responds with AI-generated reply in card format

Verify specific features:

// Test document creation
// In DingTalk: @bot create a document titled "Test"

// Test todo
// In DingTalk: @bot remind me to review code tomorrow

// Test search
// In DingTalk: @bot search for meeting notes from last week

Security Notes

• Model Risks: AI may hallucinate or execute unintended actions
• Authorization Scope: Bot operates with user permissions granted during setup
• Data Privacy: Follow company security policies if using corporate account
• Default Protections: Do NOT disable default security configs without understanding risks
• Recommended Use: Personal assistant, not production enterprise deployment

Additional Resources

• Manual Setup Guide
• DEAP Agent Integration (local device operations)
• Multi-Agent Routing Examples
• Troubleshooting Guide
• GitHub Issues
• CHANGELOG