Skip to main content
APP Capabilities

Connectors

Connect the desktop app to external services and workplace systems.

By default, QwenWork completes tasks for you within conversations. But real work doesn't happen only in the chat box — you also need to look up web pages, check your calendar, and handle email, and in the workplace you'll also use collaboration platforms such as DingTalk. Connectors are the bridge between QwenWork and these external systems: once enabled, the AI can operate your browser the way you do, read and write your system calendar and email, integrate with Microsoft 365, and complete common workflows such as To-Dos and daily reports in DingTalk using natural language.

Connectors

Without Connectors, QwenWork can only handle conversations and local files; with Connectors, it can help you get work done across apps and systems. QwenWork offers multiple ways to connect to external systems. All Connectors are configured and managed in one place under ExtensionsConnectors in the left sidebar, and fall into two main categories:
image.png

Built-in Connectors

  • Browser: Controls your browser for web operations, data extraction, and automated form filling.
  • macOS Apps: Reads and writes data from native Mac apps — Calendar, Reminders, Notes, Mail, and Contacts.
  • Microsoft 365: Reads and writes Outlook email, calendar, to-dos, contacts, OneNote notebooks, and OneDrive files.

Integration Marketplace

In the in-app marketplace, you can install and connect enterprise collaboration platforms and third-party SaaS tools with one click. Integrations in the productivity section include DingTalk, Feishu (Lark), Notion, Linear, Todoist, Canva, Supabase, Vercel, Neon, Slack, Figma, Google Calendar, Google Maps, LINE, Qichacha, and more. All Connectors are turned off by default. QwenWork gains the corresponding access only after you explicitly complete authorization and configuration.

Browser

Browser lets QwenWork control your browser — opening pages, clicking buttons, extracting data, and filling in forms.
The Browser connector currently supports only Chromium-based browsers (such as Chrome and Edge). Safari and Firefox are not supported yet.

Why you need it

Much of daily work involves repetitive operations in the browser: logging into a back-office system every day to export data, checking competitors' prices on a regular basis, or copying information back and forth between multiple pages. These tasks aren't difficult, but they are time-consuming and error-prone. Browser lets you describe what you need in a single sentence, and the AI completes the entire workflow automatically.

Core capabilities

  • Shared login state: Runs in a browser you are already signed into — no need to re-enter account names and passwords.
  • Natural-language driven: No crawler code or recorded scripts required; just describe what you want in everyday language.
  • Page awareness and interaction: The AI can "see" the page structure, understand elements such as buttons, forms, and lists, and perform clicks, typing, scrolling, selection, and other operations.
  • Data extraction and organization: Extracts structured data from web pages and outputs it as tables, Markdown, or files.
  • Multi-tab operation: Can work with multiple tabs at the same time, switching between pages and comparing data.

Quick setup

1. Install the browser extension Under Extensions in the left sidebar, click Connectors, find Browser in the recommended list, and click +. When the dialog opens with the status Waiting for extension, click Install from Chrome Web Store to complete the installation. If you run into trouble, click "Having trouble installing from the store? Try manual installation".
image.png
2. Confirm the connection Once the extension is installed and enabled, the dialog status changes to Extension connected and shows the connected browser instance — the connection is complete.
image.png
After a successful connection, the AI takes over the current tab. While a task is running, avoid operating that page manually so your actions don't conflict with the AI's.

Typical scenarios

Data collection and research
Open https://huggingface.co/models,
filter for recently updated models in the "Text Generation" category,
extract the names, download counts, like counts, and publish dates of the top 10 models,
and organize them into an Excel spreadsheet sorted by downloads from highest to lowest.
Web information extraction
Open https://github.com/trending,
extract the names, descriptions, star counts, and languages of today's trending repositories,
and organize them into a Markdown table saved to the working directory.
Industry news collection
Open Hacker News and browse today's top 10 posts,
filter for articles related to "AI" or "Machine Learning",
open each article's detail page and extract the key takeaways,
and compile them into a Markdown briefing of today's AI news.
Bulk form filling
Read "team-members.xlsx" from the working directory,
then open the company's internal admin system (https://admin.example.com),
and fill each row's name, department, and position into the new-member form.
Take a screenshot to confirm after each entry is completed.

macOS Apps

macOS Apps lets QwenWork directly read and write data from native apps on your Mac, turning your Calendar, Mail, Reminders, and more into a workbench the AI can operate.

Supported apps

Under Extensions in the left sidebar, click Connectors, find macOS Apps, and enable the following apps individually as needed:
image.png
Reminders — Reads and manages your Reminders lists. The AI can create tasks, set due dates, and mark completion status, acting as your to-do management assistant. Calendar — Reads and manages calendar events. The AI can review today's schedule, create new events, and check for time conflicts, helping you plan your day. Notes — Reads and creates notes. The AI can save important information from conversations, meeting minutes, and quick ideas directly to Notes. Mail — Reads email and creates drafts. The AI can summarize unread messages, extract key information, and draft replies, reducing the time you spend in your inbox. Contacts — Searches, reads, and creates contacts. The AI can quickly look up contact details and organize contact data in bulk.
The first time you enable a connection to an app, macOS shows a system permission request (such as "QwenWork would like to access your calendar"). Once you grant access, everything works normally and the request won't appear again.

Typical scenarios

Manage your schedule in one sentence
Check today's schedule on my calendar,
add the 3 PM meeting to Reminders with high priority,
and create a new note in Notes listing the materials I need to prepare.
Smart email summaries
Read all unread emails from today,
sort them by importance, and summarize the core content of each email in one sentence.
Save the details of the three most important emails to Notes.
If any email needs a reply today, draft one and place it in Drafts.
Cross-app information integration
Check my calendar events this week and find all external meetings.
For each meeting, look up the corresponding contact's information in Contacts,
and generate a meeting memo that includes the meeting time, attendee names, contact details, and meeting topic.

Microsoft 365

Microsoft 365 lets QwenWork directly access and manage your Microsoft account data. Once enabled, the AI can help you handle Outlook email, arrange your calendar, manage to-dos, and more.

Quick setup

Before first use, you need to authorize your account:
  1. In QwenWork, go to ExtensionsConnectors in the left sidebar, find Microsoft 365, and click Connect Microsoft Account.
image.png
2. In the browser window that opens, sign in to your Microsoft account. 3. Read the permissions request page carefully (such as "QwenWork MS365 Connector needs your permission..."). Once you confirm authorization, it is ready to use.

Supported apps

After signing in, you can enable the following services individually as needed: Mail — Reads, sends, and manages Outlook email and drafts. The AI can summarize your inbox, draft replies, or find specific emails. Calendar — Reads and manages Outlook calendar events and meetings. The AI can review your schedule and create new meeting invitations. Contacts — Reads and manages Outlook contacts. The AI can quickly find contact details for colleagues or clients. To Do — Manages personal tasks and task lists (Microsoft To Do). The AI can add to-dos and mark them complete. Notes — Accesses OneNote notebooks, sections, and pages. The AI can record meeting minutes or extract note content. Files — Browses and manages OneDrive files and Excel data. The AI can find documents in the cloud or read spreadsheet data.

Typical scenarios

Handling email and schedules
Summarize all the important emails I received today.
If any email mentions a meeting invitation, check my Outlook calendar for conflicts,
and if there is no conflict, add the meeting to my calendar and reply to confirm my attendance.
Cross-platform task management
Check my Microsoft To Do tasks for today,
find the high-priority tasks,
search OneDrive for relevant reference documents based on the task content,
and add the document links to the task notes.

DingTalk Workbench

DingTalk connects QwenWork to your organization's DingTalk: instead of clicking through multi-level menus item by item, you can complete workflows such as creating To-Dos, submitting or drafting reports, and handling attendance and calendar events in natural language (the exact scope depends on what your enterprise administrator has enabled).

Why you need it

In everyday work, handling matters in DingTalk usually requires tedious steps: finding the right app, clicking into the details, filling in a form, and submitting it. The Connector lets QwenWork connect directly to the DingTalk Workbench (DWS), compressing these "multi-click" operations that span multiple pages into "one-sentence" natural-language instructions — keeping you in your current work context.

Core capabilities

Subject to what your enterprise administrator has enabled, the Connector can operate more than ten DingTalk modules through a large language model. Typical capabilities include:
  • To-Do management: For example, type "finish reviewing the three tables by 8 PM today", and the AI automatically parses the time and content and creates a To-Do. You can also view your To-Do list, change completion status, or update goals directly in the conversation window, without leaving the editor.
  • Report filling: Say goodbye to manual compilation. QwenWork automatically gathers your To-Dos and the work you did during the day (such as submission records and task progress), organizes them into a cleanly formatted report, and pushes it to DingTalk with one click. Weekly and monthly reports work the same way — no page switching or opening report templates required.
  • Attendance and calendar: Within the scope authorized by the Connector, initiate attendance check-ins, query shift schedules, or book and manage calendar events.
  • Approval workflows: Subject to what your enterprise has enabled, quickly initiate or handle various approval flows (the specifics depend on the configuration on the DingTalk side).
The modules and permissions actually available depend on your enterprise's configuration on the DingTalk side and the capability scope granted to the app.

Quick setup

1. Open the DingTalk Connector Under ExtensionsConnectors in the left sidebar, switch to the Marketplace tab, find DingTalk, and click the card. 2. Sign in to your DingTalk account The Sign in to DingTalk dialog appears. Read the permission description, then click Sign in to DingTalk at the bottom to complete authorization.
image.png
3. Confirm the installation Once authorization is complete, view or manage the connector under Installed. After completing the steps above, you can operate To-Dos, reports, attendance, calendar, approvals, and more in natural language (subject to the capabilities your enterprise has enabled).

Typical scenarios

Create a DingTalk To-Do
Create a To-Do in DingTalk for 8 PM today, titled "Review of three tables".
Draft a daily report
Based on yesterday's meeting notes and completed To-Dos, list the key points as a draft so I can paste it into today's DingTalk daily report.
You can turn frequently used DingTalk instructions into a Skill, or combine them with Scheduled Tasks for periodic automation.

Custom MCP Services

MCP (Model Context Protocol) is an open protocol that standardizes how AI applications connect to external tools and data sources. Think of it as the "USB port" of the AI world — just as USB makes all kinds of peripherals plug-and-play, MCP lets AI call a wide variety of external services in a unified way. Before MCP, AI assistants could only answer questions based on the conversation context. With MCP, QwenWork can directly operate your calendar, query documents, fetch map information, and more — truly becoming an assistant that can "take action".

How it works

MCP uses a client-server architecture:
  • MCP client: QwenWork itself, responsible for initiating requests and presenting results.
  • MCP server: Deployed by service providers (such as DingTalk and AMap), responsible for executing the actual operations and returning data.
When you make a request in QwenWork, the AI automatically determines whether an MCP service needs to be called, completes the invocation through the protocol, and presents the result to you — all you need to do is describe what you want in natural language, as usual.

Add an MCP service

Open the QwenWork desktop client and go to ExtensionsConnectors in the left sidebar. Click + Add in the upper-right corner of the page, and choose one of the following two methods:
Product interface

Paste JSON configuration

Best for cases where you copy a JSON configuration directly from an MCP marketplace. Click + AddPaste JSON configuration; the Import via JSON dialog appears — paste the configuration and click Import.
Product interface
Example JSON configuration format:
{
  "mcpServers": {
    "service-name": {
      "type": "streamable-http",
      "url": "https://mcp-server-url.example.com/sse",
      "headers": {
        "Authorization": "Bearer your-token"
      }
    }
  }
}

Fill in the configuration manually

Best for cases where you fill in the service configuration by hand. Click + AddFill in configuration manually. In the "Add MCP Server" dialog that appears, first choose a type from the Server type dropdown. Different types correspond to different configuration forms:
Product interface
Streamable HTTP / SSE
Remotely hosted MCP services, accessed via URL. Most services offered by MCP marketplaces are of this type. Configuration is simple, and this is the recommended first choice.
  • Server name: Give the service an easily recognizable name.
  • Server address: Paste the URL copied from the MCP marketplace.
  • Headers (optional): If the service requires authentication, click + Add Header to add request headers (for example, Authorization: Bearer your-token). You can also click Paste to import them in bulk from the clipboard.
Streamable HTTP is the upgraded version of SSE, with more powerful functionality. If the MCP marketplace provides a Streamable HTTP URL, simply choose the Streamable HTTP type.
STDIO
MCP services that run locally and are launched from the command line. Suitable for developers using local tools or command-line integrations; the corresponding runtime environment (such as Node.js or Python) must already be installed on your machine.
  • Server name: Give the service an easily recognizable name.
  • Command: Paste the complete launch command, such as npx -y @modelcontextprotocol/server-filesystem.
  • Environment variables (optional): Environment variables the service needs at runtime (such as API keys and access tokens). Click + Add Variable to add key-value pairs one by one, or click Paste to import them in bulk from the clipboard.
Once you've filled everything in, click Add.
How do you choose between pasting JSON configuration and filling in the configuration manually? Pasting JSON configuration automatically parses all fields in the configuration (type, address, authentication, etc.) in one step; filling in the configuration manually requires entering each item but is more flexible. Both methods produce the same result — choose whichever is more convenient.

Manage MCP services

Open the Connectors page and switch to the Installed tab. Added MCP services appear in the Custom section. You can:
  • Enable or disable a service with its toggle.
  • Expand a service to view the list of tools it provides.
  • Click the edit icon to modify its configuration, or click the delete icon to remove the service.

Using Connectors Together

When multiple types of Connectors are enabled at the same time, QwenWork can collaborate across apps within a single task: Browser + Calendar
Check all the meetings on my calendar tomorrow.
For each meeting, use the browser to search for background information on the attendees' companies,
and generate a pre-meeting preparation briefing saved to Notes.
Browser + Mail
Read all customer emails received today.
For each product issue mentioned in the emails, use the browser to search our documentation site for the corresponding solution,
and organize them into a comparison table of issues and solutions saved to the working directory.
Full-feature collaboration
Please do the following:
1. Check today's calendar and list all meetings
2. Read unread emails and filter for content marked as important
3. Open the browser and check today's To-Dos on the team project board
4. Consolidate the information above into a daily work briefing, save it to Notes, and create a corresponding reminder
Scheduled Tasks
Combine these multi-connector tasks with Scheduled Tasks to achieve periodic automatic execution. For example, set the "daily work briefing" above to run at 9 AM every workday, and the briefing will be ready by the time you arrive at the office.

Best Practices

Give an explicit URL or operation path

When using the Browser connector, providing the target URL directly in your prompt is faster and more accurate than letting the AI search for it on its own. For example, say "Open https://analytics.google.com" rather than "Find Google Analytics for me".

Combine Connectors with file reading and writing

Connectors can work together with QwenWork's local file capabilities. You can have the AI scrape data from a web page and save it as a local Excel file, or read a local document and then bulk-fill forms on a website.

Run complex workflows through manually once first

For complex workflows that involve multiple page jumps, we recommend first guiding the AI through the entire process in a regular conversation. Once the steps are confirmed, solidify the prompt into a Skill or Scheduled Task for reuse.

Make good use of human-AI collaboration

When you encounter steps the AI can't complete on its own, such as CAPTCHAs or QR-code sign-ins, you can take over at any time. After you complete the verification manually, the AI continues with the rest of the task in the signed-in state.

FAQ

After enabling a Connector, the conversation says the related tools can't be found? Right after enabling or disabling a Connector, you need to create a new conversation task for it to take effect. Please don't continue asking in the original conversation. What if I encounter a sign-in page or CAPTCHA? QwenWork can't automatically pass complex image CAPTCHAs, SMS verification, or QR-code sign-ins. When the AI navigates to such a page, simply complete the verification manually, and the AI will continue once sign-in succeeds. What if a macOS app permission was denied? If you clicked Deny when the dialog first appeared, you need to grant the permission manually: open System Settings → Privacy & Security, find the corresponding permission category (Calendar, Reminders, etc.), and add QwenWork to the allowed list. Is it normal for browser operations to be slow? The Browser connection needs to perceive page content in real time, which involves screenshots and element analysis, so it is slightly slower than pure conversation. Network conditions and page complexity both affect execution time. For data-heavy tasks, we recommend describing the steps in stages in your prompt, letting the AI complete them step by step. What if an MCP doesn't work after being added?
Please check the following in order:
1. Confirm that the JSON configuration or URL was copied completely and is correctly formatted.
2. Confirm that your network connection is working normally (some MCPs require a specific network environment).
3. Confirm whether the corresponding service requires additional authorization or sign-in.
4. Try removing the service under **Connectors** → **Installed** and adding it again.
What if MCP service calls are delayed? Response times for MCP services depend on the service provider. Most services return results within a few seconds. If timeouts persist, check your network environment or contact the service provider.
Connectors - QwenWork