For most users, the moment OpenClaw becomes real is not when installation finishes. It is the first time a message sent from a real chat app gets a real reply back.
01 Channel Access Overview
OpenClaw can manage 20+ messaging platforms through one general pattern:
Create credentials -> write config -> start Gateway -> complete pairing
Common platforms and complexity
| Platform | Type | Difficulty | Typical time |
|---|---|---|---|
| Telegram | Built-in | Very low | 5 minutes |
| Plugin / official support | Very low | 5 minutes | |
| Discord | Built-in | Low | 15 to 20 minutes |
| Feishu | Built-in / plugin | Low to medium | 15 to 20 minutes |
| Built-in | Medium | 10 to 15 minutes | |
| Slack | Built-in | Medium | 25 to 40 minutes |
| DingTalk | Community plugin | Medium | 20 to 30 minutes |
| WeCom | Community plugin | Medium | 20 to 30 minutes |
| Signal | Built-in | Medium | 20 to 30 minutes |
| iMessage | Extension | Medium to hard | 30 to 45 minutes |
Recommended order for beginners
If your goal is just to prove the full loop works:
TelegramQQDiscordorFeishuDingTalk,WeCom, orWhatsApp
02 International Platforms
Telegram: the easiest place to start
Telegram is the cleanest onboarding channel because it supports long-polling and does not require a public IP or reverse proxy.
Typical flow:
- Find
@BotFather - Run
/newbot - Create the bot and get the token
- Write the config and restart Gateway
- Send the bot a message and complete DM pairing
Example:
channels:
telegram:
enabled: true
botToken: "YOUR_BOT_TOKEN"
dmPolicy: pairingDiscord: good for teams and communities
Discord is still straightforward, but there are more steps:
- Create an application
- Create a bot
- Enable
Message Content IntentandServer Members Intent - Invite the bot with OAuth2
- Get Server ID and User ID
- Pair through DM
WhatsApp: popular, but session hygiene matters
WhatsApp uses QR-based connection through Baileys rather than the official Business API.
Typical flow:
- Run
openclaw onboard - Choose WhatsApp
- Scan the QR code from your phone
Practical guidance:
- use a dedicated number if possible
- treat session credentials like passwords
- expect occasional re-pairing
- prefer
NodeoverBunfor this scenario
Slack: common in business environments
Slack often runs in Socket Mode, which avoids needing a public callback URL.
You usually need:
- an
xapp-App-Level Token - an
xoxb-Bot Token - a proper set of bot scopes
Because OpenClaw can execute real commands, it is safer to run Slack-connected agents on a dedicated VM or server rather than your primary workstation.
Signal and iMessage
Signal relies on signal-cli. iMessage usually relies on BlueBubbles.
For iMessage, two things matter most:
- a Mac must stay online
- webhook access should be protected with authentication
Example BlueBubbles config:
extensions:
bluebubbles:
enabled: true
serverUrl: "http://localhost:1234"
password: "YOUR_PASSWORD"03 China-Focused Platforms
QQ: one of the easiest local entry points
QQ Bot setup is fast, QR-driven, and relatively friendly to media-rich messages.
Typical steps:
- Register as a QQ Bot developer
- Create the bot
- Get App ID and Token
- Write config
- Start using private chat or group scenarios
Feishu: a natural fit for team workflows
Feishu support is a strong path for:
- internal assistants
- team notifications
- document and group workflows
A common path is:
- Create a self-built app in the platform
- Get
App IDandApp Secret - Run
openclaw onboard - Choose Feishu and fill in credentials
DingTalk and WeCom
These are still mostly community-plugin driven, but already fairly mature.
DingTalk
Stream mode is generally the better option because it uses a long-lived WebSocket connection and avoids public callback plumbing.
WeCom
WeCom usually appears in two patterns:
- Agent mode
- Bot mode
It works well for internal automation and has already been validated by several cloud deployment stacks.
Personal WeChat: high demand, high complexity
This area should be approached carefully. There is no official bot API, which means every route carries maintenance and account-risk tradeoffs.
Common approaches include:
WeCom relayiPad-protocol relayMini-program integration
General advice:
- do not use your main account
- do not treat it as permanently stable
- expect protocol breakage over time
openclaw-china
If you want multiple mainland channels at once, the unified plugin package saves time.
git clone https://github.com/BytePioneer-AI/openclaw-china.git
cd openclaw-china
pnpm install
pnpm build
openclaw china setup04 Remote Access and Multi-Device Use
By default, Gateway listens only on ws://127.0.0.1:18789. If you want outside devices or services to reach it, you need explicit remote-access setup.
Tailscale Serve / Funnel
This is one of the most practical approaches.
tailscale serve --bg https+insecure://127.0.0.1:18789
tailscale funnel --bg https+insecure://127.0.0.1:18789| Mode | Scope | Best for |
|---|---|---|
| Serve | Devices inside your Tailscale network | Access from your own phone or tablet |
| Funnel | Public internet | Webhook-style integrations |
SSH forwarding
If OpenClaw runs on a remote server:
ssh -L 18789:127.0.0.1:18789 user@your-server
ssh -fNL 18789:127.0.0.1:18789 user@your-serverDashboard and clients
Once Gateway is running, the Web UI can help you inspect:
- session state
- model settings
- channel status
- token usage
openclaw gateway --port 18789 --verboseThere is also a macOS menu-bar client, and iOS / Android client work has already appeared in the repo.
Best first path
For most people the smoothest route is:
Telegram or QQ -> private pairing works -> add Feishu / DingTalk / WeCom -> then solve remote access