2026 OpenClaw Telegram and WhatsApp channels on MacLogin cloud Mac: pairing, dmPolicy, and probe troubleshooting

Teams that already finished openclaw onboard --install-daemon on a MacLogin mini often hit a different wall: openclaw gateway status is green, Control UI loads over SSH forwarding, but Telegram DMs stay silent and WhatsApp groups never trigger the agent. This 21 May 2026 runbook explains how to configure Telegram botToken + dmPolicy, relink WhatsApp QR sessions, and use openclaw channels status --probe with logs—not guesswork. You get a policy matrix, channel-specific steps, a symptom triage table, and region/proxy baselines for Hong Kong, Japan, Korea, Singapore, and United States nodes. Start from onboard + daemon bootstrap, add health curl monitoring, and use stderr triage when probes fail mysteriously. Links: help, pricing, VNC.

Who needs messaging channels on a leased gateway—and what “connected but mute” really means

The reader runs OpenClaw as an always-on agent front door for ops or revenue teams: alerts should land in Telegram, account managers want WhatsApp group mentions, and engineering still SSHes into the same Mac mini. The failure mode is subtle—channels register in config, yet human senders see no replies because pairing queues, mention rules, or stale webhooks block delivery.

• Pairing backlog: first DM requires openclaw pairing list approval; without it, policy rejects silently.
• Token loading race: referencing ${TELEGRAM_BOT_TOKEN} without the variable present at gateway start yields 401 at getMe.
• WhatsApp QR drift: Baileys sessions expire; operators confuse “linked yesterday” with “valid now.”
• Group mention gates: requireMention defaults leave groups quiet unless someone @-tags the bot.
• Plugin/config clutter: recent regressions show empty channel lists when experimental plugins blocks channel init—doctor may suggest unsetting them.

Channel policy matrix: dmPolicy, pairing, and group mention rules

Pick policies deliberately per environment; mixing production strictness with staging permissiveness on one gateway creates “works for Alice, not Bob” tickets.

Channel	Control knob	Strict (prod)	Permissive (staging)	Common mistake
Telegram DM	dmPolicy + pairing	Pairing required; approve per sender	Allowlist test bot chat only	Token valid but pairing never approved
Telegram groups	Privacy mode / mentions	Bot must be @-mentioned	Disable privacy via BotFather for lab	Expecting replies without mention
WhatsApp DM	Session + allowFrom	Explicit allow list	Single operator phone for pilot	Stale QR after lease handoff
WhatsApp groups	requireMention	Mention required	Temporarily off for demo	High-volume group spam when off

Prerequisites: healthy gateway before any channel token

Channels are not a shortcut around a broken daemon. Confirm:

• openclaw gateway status reports runtime running and RPC probe OK.
• Loopback listener on 18789 matches your SSH tunnel recipe from the onboard article.
• Outbound HTTPS from the mini reaches api.telegram.org and WhatsApp web endpoints—test with curl -I from the host, not your laptop.
• Node semver pinned in CMDB (commonly 22.14+, newer gateways on 24.x).

Telegram: bot token, webhook cleanup, and pairing approval

Create a bot via @BotFather, copy the token, and place it in gateway channel config as plaintext during debugging if env injection races caused empty startup. If you previously used webhooks on the same token elsewhere, clear stale state:

curl "https://api.telegram.org/bot<TOKEN>/deleteWebhook"

Restart the gateway, then run openclaw channels status --probe. Send /start from a test account; if policy requires pairing, approve via openclaw pairing list telegram and the documented approve subcommand. Capture the approval event ID in your change ticket.

WhatsApp: QR relink, proxy env, and group mention drills

WhatsApp uses the gateway-linked web session. Open Control UI over your SSH tunnel or VNC per VNC guidance, start linking, scan QR within 60 seconds. If login returns 408, set reachable HTTPS_PROXY for the mini’s region and add NO_PROXY exceptions only when security approves.

After link, run a DM canary, then a group message that @-mentions the bot if requireMention is enabled. Random disconnect loops: openclaw gateway restart, relink if probes still fail.

Nine-step channel rollout from zero to verified replies

1. Freeze a 20-minute window and announce in the channel covering HK/JP/KR/SG/US operators.
2. Verify gateway health with status + curl probe runbook—not only UI gut feel.
3. Write Telegram token into config; run openclaw doctor before restart.
4. Delete stale Telegram webhook if migrating from another host.
5. Restart gateway via launchd kickstart or approved reload script.
6. Run channels status --probe and archive JSON output to the ticket.
7. Approve pairing for the first DM sender; screenshot policy screen for auditors.
8. Link WhatsApp QR; store session path and lease ID in CMDB.
9. Run group + DM canaries; tail logs for 10 minutes before declaring success.

Probe triage matrix when status looks fine but users hear nothing

Symptom	Probe / log signal	Likely cause	First fix
Channel missing from openclaw status	No telegram/whatsapp rows	Plugin block or config not loaded	openclaw config unset plugins; restart
Telegram 401 at startup	getMe returned 401	Bad token or env race	Inline token; regenerate via BotFather
DM ignored	Pairing pending in list	dmPolicy	Approve sender or relax staging policy
Group silent	No mention events in logs	requireMention	@ bot or adjust policy temporarily
Replies minutes late	Stale TUI clients in doctor	Client backlog	openclaw doctor --fix
WhatsApp disconnect loop	Probe flaps connected/disconnected	Proxy or session corruption	Fix proxy; relink QR; restart gateway

Region, proxy, and DNS baselines on MacLogin metros

Place the gateway near users and near messaging APIs. Illustrative RTT targets from a well-routed corporate network (measure yours):

MacLogin region	Typical use	Watch-out
Hong Kong (HK)	Greater Bay ops bots	Cross-border proxy rules affecting WhatsApp web
Japan (JP)	Tokyo daytime support	IPv6 preference quirks—test curl -4
Korea (KR)	Seoul engineering alerts	Corporate MITM replacing certs—pin carefully
Singapore (SG)	ASEAN hub groups	Low RTT but strict egress allowlists
United States (US)	Americas customers	Long Pacific path if users still in APAC

Document HTTPS_PROXY, HTTP_PROXY, and NO_PROXY in the lease record when channels depend on them.

FAQ: channels vs hooks, multi-gateway, and security

Are channels the same as webhooks? Related but distinct—channels maintain messenger transports; HTTP webhooks are separate ingress. You may use both; do not disable one while debugging the other unless intentional.

Can I run two gateways on one mini? Avoid port 18789 collisions; prefer two leases or distinct loopback ports with explicit SSH maps.

Is storing botToken in config safe? Treat the mini as a secrets host: restrict SSH, rotate tokens on offboarding, prefer vault injection once env races are solved.

Does VNC help? Yes for QR scan UX; pairing still lives in CLI/UI commands.

Why Mac mini M4 on MacLogin fits always-on messaging gateways

Channels need a stable Node process, persistent session files, and outbound HTTPS—not a laptop that sleeps. Apple Silicon M4 minis keep thermals steady under polling loops, and MacLogin’s HK/JP/KR/SG/US footprint lets you place the gateway beside both users and APIs. Renting means you can spin a dedicated “channels only” mini with isolated WhatsApp sessions instead of overloading a developer’s daily driver machine.

Compare core counts and regions on pricing, and keep help open for SSH/VNC access patterns while you link messengers.