# WeCom Webhook Sending Use an internal WeCom group's Message Push webhook (formerly called a group robot) to send notifications and task results. This needs only the group's webhook URL, without Bot ID/Secret, application credentials, a long connection or a public inbound server. For conversations with an agent, use the [WeCom intelligent robot](./wecom); for a self-built application, see [WeCom Callback](./wecom-callback). ## Create a group destination 1. Open an internal WeCom group's settings and find Message Push. 2. Create a custom Message Push and copy its webhook URL, beginning with `https://qyapi.weixin.qq.com/cgi-bin/webhook/send?key=`. 3. Check your organization's creation permissions if the entry is unavailable. Administrators can restrict which members may create Message Push, and owner-managed groups can restrict the entry to the group owner. External groups are not supported by this setup. The official creation conditions describe internal-group and administrator permissions; they do not establish a universal enterprise-certification requirement or exemption. The URL contains the sending credential. Keep it in Channel configuration or an environment variable, never in a task prompt or delivery target. Use the URL as issued by WeCom. Local validation allows exactly one `key` parameter containing letters, digits, `-`, `.`, `_` or `~`; percent encoding, fragments and additional parameters are not accepted. There is no separate signing secret. ## Configure a Channel Configure this integration in [Web Shell](#use-web-shell) or `settings.json` as shown below. Add a named instance to the workspace's Channel settings: ```json { "channels": { "team-notices": { "type": "wecom-webhook", "webhookUrl": "$WECOM_WEBHOOK_URL" } } } ``` Set `WECOM_WEBHOOK_URL` in the environment used to start Qwen Code, then run the daemon in that workspace: ```bash qwen serve --channel team-notices ``` One named instance represents one group. Create additional instances for other groups. Changing an instance's URL changes future deliveries through it. Starting makes the Channel locally ready to send. It does not send a test message or verify the remote credential. Stopping cancels in-flight sends; restart or reload applies saved configuration changes. ## Use Web Shell If your build contains the backend support but predates the Web integration, Channels does not list this method or existing `wecom-webhook` instances. Use the `settings.json` and `qwen serve --channel` steps above until you upgrade to a build containing the Web integration. In Channels, choose WeCom and then Webhook (send only). Select the workspace, save the webhook URL and start the Channel. No signing secret is needed. Receiving, pairing and agent-session settings do not apply to this method. The intelligent-bot method remains under WeCom; WeCom Callback keeps its separate entry. In Scheduled Tasks, choose a persistent task session and select this Channel as the delivery destination. Web Shell fills its configured group automatically; no observed contacts or manually entered chat ID are needed. Existing per-run and manual-trigger delivery restrictions remain in effect. The same controls are available through `WebShellWithProviders` when the host enables Channels and Scheduled Tasks and supplies a daemon workspace. ## Explicit API delivery Notify and explicit prompt/scheduled delivery use the existing Channel target. This adapter owns one logical destination, `chat/default`: ```json { "text": "# Daily report\n\nThe checks passed.", "delivery": { "kind": "channel", "target": { "channelName": "team-notices", "type": "chat", "id": "default" } } } ``` This is a workspace Notify request body. `default` names the configured group; it is not a WeCom chat ID. Other IDs, user targets and threads are rejected. Tasks store the logical destination; the credential stays in Channel settings. Persistent scheduled tasks use the same destination. Delivery retains the existing best-effort successful-final-result behavior, including current per-run and manual-trigger restrictions. ## Message formatting and limits - Sends `markdown_v2` with the original text. Its supported subset includes headings, lists, links, fenced code and tables; ordinary text, Chinese and emoji are valid input. Complete GitHub-flavored Markdown compatibility is not guaranteed. - This format does not support mentions or font-colour markup. The adapter sends no mention arrays and does not turn task text into mentions. Mention-looking strings such as `<@all>` remain in the body. - Clients before 4.1.36 (Android before 4.1.38) display the message as plain text. Use a current client for Markdown rendering. - The final **content** is limited to **4096 UTF-8 bytes**, before JSON escaping. Chinese and emoji take more bytes than ASCII. This is not a 4096-character limit or a limit on the complete JSON request. - Oversized content fails explicitly before sending. There is no silent adapter truncation, automatic splitting or retry. The shared delivery bound still applies before the platform-specific check. - WeCom permits **20 messages per minute per webhook**. Keep scheduled reports concise and avoid bursts; throttled deliveries report failure without queuing. ## Failures and transport Both HTTP status and the numeric WeCom `errcode` are checked. Invalid or disabled credentials, rejected group destinations and rate limits produce observable failures. Diagnostics omit the URL, key, message content and raw provider error text. Each attempt makes at most one request with a 15-second timeout. API acceptance is not a read receipt; a timeout or cancellation can leave remote acceptance unknown. Transport honors the resolved workspace proxy and existing `NO_PROXY` behavior. This mode does not receive messages, run inbound webhook tasks or start Channel loops. It adds no attachments, interactive cards, automatic failure/cancellation alerts, durable retries or generic webhook protocol. See the official [Message Push protocol](https://developer.work.weixin.qq.com/document/path/91770) and [setup guide](https://open.work.weixin.qq.com/help2/pc/14931) for current platform requirements.