Tencent WeChat ClawBot 2026:
Official OpenClaw Install, Version Gates, and Eight Guardrails

Domestic teams rarely ask for “another Telegram bot.” They want lightweight commands inside the app they open dozens of times per day: research, drafting, skill triggers, short automations. Unofficial bridges fail in three predictable ways:

• Account risk: simulated clients and protocol hacks can trigger platform enforcement that also affects business contacts.
• Unowned uptime: login flows, captchas, and upstream API changes become midnight script work with no vendor SLA.
• Product mismatch: multi-bot Telegram patterns do not map cleanly to WeChat routing; group features amplify state bugs.

ClawBot uses Tencent’s plugin system: you discover it under Settings → Plugins, run the published CLI on the gateway host, and authorize with a phone QR scan. Traffic still passes Tencent security review—stability improves, but you must accept moderation, direct-chat-only scope, and some delivery limits as product rules, not installation defects.

Operations teams should document three ownership lines before go-live: who upgrades WeChat on test phones, who rotates model API keys on the gateway, and who approves skill installs that can touch the filesystem. Without that split, the first production incident becomes a blame game between “WeChat broke” and “OpenClaw broke” when the real issue is an unaudited skill writing outside its workspace.

Treat ClawBot as an official channel plugin, not a generic IM SDK: install is fast; version gates, moderation, and gateway placement are the real acceptance tests.

Individual builders should default to ClawBot. Move to the WeCom plugin when you need org-wide distribution and audit—not both on the same experimental gateway profile. Mixing tokens and routing tables between the two paths is a common source of “wrong agent answered” bugs in staging.

ClawBot binds to your OpenClaw instance—not a generic cloud model inside WeChat. Before install:

• WeChat client: iOS commonly requires 8.0.70+; Android gates follow the plugin page and may still be gray-release by region.
• Device: binding uses mobile WeChat; desktop WeChat cannot finish authorization.
• CLI host: run install in a shell on the Mac where the gateway runs (local or SSH into CALMVPS bare metal).
• Networking: phone and CLI host must complete the QR handshake; split latency between model API egress and WeChat separately.

Topology A: laptop gateway + local CLI + phone QR. Topology B: gateway on rented bare-metal Mac 24/7, SSH in to run CLI, same phone QR—better when the laptop sleeps.

npx -y @tencent-weixin/openclaw-weixin-cli@latest install

Verify package name and README on npm before production; if Tencent updates flags, copy the command from the plugin detail page.

1. Upgrade WeChat: Settings → About, meet the plugin requirement, update from the store if needed.
2. Open Plugins: Settings → Plugins → WeChat ClawBot → copy the install command.
3. Confirm gateway health: run status checks; on remote nodes, SSH first.
4. Run CLI: paste npx -y @tencent-weixin/openclaw-weixin-cli@latest install and wait for the QR.
5. Authorize on phone: scan with mobile WeChat; you bind this gateway instance, not a shared vendor model.
6. Validate channel: confirm WeChat channel online in the control UI; if missing, use one-click update and hard-refresh the browser (Cmd+Shift+R), restart gateway if needed, re-scan.

Inbound messages route through OpenClaw’s resolveAgentRoute by channel, accountId, and peer. For most individuals, one WeChat account + a primary agent that delegates internally beats switching agents inside the same thread.

Playbook B—multiple WeChat accounts mapped to different agents—is supported when you need hard boundaries (e.g., “personal assistant” vs “ops bot”). Each account gets its own peer context; do not expect Telegram-style multi-bot tabs inside one chat. After binding, send a harmless ping (“who are you / which model”) to confirm routing before enabling skills that shell out or touch browsers.

If you host the gateway on CALMVPS bare metal, run the CLI over SSH in the same user session that owns launchd for ai.openclaw.gateway. Closing your laptop should not stop WeChat reachability; only gateway health should.

• Direct chat only: do not expect group @mentions; broadcast via WeCom or other channels.
• Content moderation: Tencent filters sensitive domains (e.g., DeFi, wallets, crypto wording)—empty replies may be policy, not model failure.
• Spare account: bind a secondary WeChat when OpenClaw has broad system permissions.
• Workspace boundaries: tighten writable paths and system.run approvals.
• Idle sessions: long silence may require a new user message to wake the channel.
• Android gray release: missing plugin entry usually means version/region, not “install harder.”
• File delivery: processed artifacts may need email or drive handoff, not WeChat attachments.
• Missing channel row: align gateway/CLI versions, check logs after UI refresh.

When debugging “silent failures,” capture gateway logs and the exact user utterance. Moderation often returns empty bodies without a stack trace; comparing the same prompt on Telegram or Discord helps prove the model path is fine while the WeChat channel is filtered.

Tencent’s developer article documents the official ClawBot launch timeline and screenshots—re-open the link after publish to confirm steps have not changed.

WeChat official ClawBot: one command to connect OpenClaw

• WeChat plugin gate (iOS): public tutorials cite 8.0.70+.
• Official CLI package: @tencent-weixin/openclaw-weixin-cli via npx … install.
• Default gateway control port: 18789 (same as remote dual-machine docs).
• Remote sizing: multi-channel + cron + WeChat entry: start at M4 24GB; heavy local models or skill trials: M4 Pro; log growth: 1TB/2TB before CPU upgrades.

Keeping the gateway on a sleeping laptop breaks 24/7 WeChat reachability and token hygiene. Generic Linux VPS lacks macOS toolchain and TCC-related node workflows. A time-shared virtual Mac farm may cap custom launch agents or share disk IO—fine for demos, painful when WeChat is your on-call pager.

Reference workflow: gateway + WeChat channel on a CALMVPS Singapore monthly M4 24GB node; local MacBook stays a Remote node for screen capture and Xcode side tasks; burst compile jobs land on a weekly parallel worker without touching gateway tokens. WeChat sessions stay on the hub; build spikes absorb on the worker.

For teams that need a stable gateway, official WeChat entry, and reproducible triage, CALMVPS multi-region bare-metal Mac is usually the better host: dedicated Apple Silicon, ~120s delivery, daily/weekly parallel workers for build spikes without upgrading the hub. See the CALMVPS pricing page.