Skip to main content
Skills provide reusable work methods; connectors let Waker call external applications or MCP tools; IM channels let external conversations deliver messages to Waker. Each serves a different purpose:
Choose the right entry point: Install work methods under “Skills”; configure external tools and DingTalk DWS under “Connectors”; pair external conversations under “IM.”

Installing and Managing Skills

Installing from the Skill Marketplace

  1. Go to the target Waker → “Skills” → “Skill Marketplace.”
  2. Search for and open the target Skill; after confirming its purpose and source, click “Install.”
  3. Go to “My Skills” to confirm the status is “Installed.”
Only install Skills directly related to the Waker’s responsibilities to avoid conflicts between similar capabilities.

Uploading Custom Skills

  1. On the “Skills” page, click “Upload Skill” and select a .md file or a supported archive.
  2. The .md file must contain name and description; archives must contain SKILL.md.
  3. After uploading, go to “My Skills” to confirm the result.
Do not upload Skills from unknown sources or ones that read sensitive directories. Check SKILL.md and its referenced content before uploading.

Post-Change Verification

To roll back or uninstall, go to “My Skills” → target Skill → “More Actions.” Before making changes, check whether any automated tasks or WakerFlows depend on the Skill. After changes, start a new conversation and verify with a minimal task. Completion criteria: The current version or installation state matches expectations; capability changes are effective in new conversations; workflows that depend on the Skill do not produce missing-capability errors. If installation or upload fails, check the network, file structure, and same-name versions, then retry.

Adding Connectors and MCP Services

Connectors manage external applications and MCP services. Go to Waker details → “Connectors,” then choose an addition method based on the configuration source:

Adding a Custom MCP

  1. In “Applications,” click “Add.”
  2. When you only have information for a single service, select “Manual Configuration” — fill in the name, startup command or server address, and optionally set environment variables, Headers, and timeout.
  3. When you have a complete configuration, select “Paste JSON Configuration” — paste the complete JSON containing mcpServers, confirm the page correctly identifies the service, then click “Import.”
  4. Before saving, confirm the command or URL is reachable, expose only necessary directories, and store credentials in protected environment variables.

Authorize and Verify

After adding or importing, follow these steps in order:
  1. When the status shows “Authorization Required,” complete authorization.
  2. Click “Detect” or “Re-detect” and wait for the status to become “Connected.”
  3. Refresh the tools and enable only the tools the current Waker needs.
  4. Start a new conversation and verify with a minimal task.
If the status is not “Connected” or the tool count is 0, re-check the configuration, authorization, and startup command, then run detection again.

Enabling and Verifying DingTalk DWS MCP

DingTalk DWS MCP is a “System” connector that can read, summarize, and send DingTalk group messages within its authorized scope.
  1. Go to the target Waker → “Connectors” → “System” and find “DingTalk DWS.”
  2. Follow the page prompts to install, update, or authorize the DWS CLI.
  3. Refresh the page and confirm the connector status is “Connected.”
After authorization, verify in a regular conversation first: read a small number of messages from the test group, generate a summary, and send fixed text to the designated test group. Completion criteria: DWS tool calls succeed, read and send targets are correct, and summaries match the original messages. Before enabling automatic message sending in production, limit the target group and sending frequency, and specify “do not send on error” in the task description. After modifying connectors, open a new conversation to test.

Configuring IM Channels and Session Pairing

The main navigation “IM” is used to manage channels and sessions centrally; the “IM” in Waker details shows conversations currently handled by that Waker.

Adding a Channel

  1. Go to “IM” → “Channel Configuration” → “Add Channel.”
  2. Select the channel and complete authorization or fill in credentials via the wizard.
  3. Set the pairing mode, responding Waker, model, and working directory; use pairing mode preferentially in production environments.
  4. Save and confirm the status is “Connected.”
Available channels depend on the current account and page.

Processing Pairing Requests

  1. Go to “IM” → “Session Management” → “Pending Pairing Requests.”
  2. Verify the session name, initiator, channel, and type.
  3. Select the responding Waker, model, and working directory, then click “Allow”; for untrusted requests, select “Ignore.”
  4. Confirm the status is “Active” in “Paired Sessions.”
You can also use “Add Pairing” to manually select a group chat or direct message.
Qmate pairing note (applies to Qmate template only): Qmate processes messages by default through a preset Q&A WakerFlow — the pairing page does not require selecting a workflow. Node models are configured in the WakerFlow; for direct messages, prioritize filling in the Open ID. After creating a pairing, confirm the channel is enabled and check Q&A records and execution records in the test group.

Verifying the Message Pipeline

  1. Confirm the channel is connected and the target Waker is running.
  2. Send a test message from a paired session.
  3. Check the Waker, model, and working directory used for the reply; verify non-sensitive attachments as needed.
If messages or pairing are abnormal, check the channel connection, pairing status, responding Waker, and working directory. For Qmate, also check the preset Q&A workflow. The “DingTalk User” proxy-send mode sends AI replies under your own identity. Before enabling, strictly limit the scope of monitored conversations and take responsibility for the auto-reply content.

Pre-Launch Integration Checklist

  • Enable Skills, connectors, and IM channels with minimum privileges — verify read-only capabilities first, then open write or message-sending capabilities.
  • Store Tokens, Cookies, AppSecrets, and Personal Access Tokens in protected credentials or environment variables — never write them into body text, logs, or screenshots.
  • Before deleting or modifying an integration, check dependencies first; after changes, perform a minimal verification in a new conversation.