ClawX OpenClaw: Skills, Channels, And Models In Detail

ClawX OpenClaw is the new free desktop app for OpenClaw, and three of its features are worth focusing on if you want to get serious value out of it rather than just using the chat UI.

Most ClawX walkthroughs cover the install and the chat UI. This post is different. I'm going deep on the three features I use every single day — skills management, channel integrations, and provider switching.

If you're going to use ClawX OpenClaw seriously, master these three.

Why These Three Matter

The chat UI is nice, but the chat UI is just the surface. The real power is in the skills your agent can use, the channels it talks to people on, and the models it runs on.

Get those three right and you've got an actual agent doing actual work.

Skills In ClawX OpenClaw

Skills are what give your agent superpowers. Without skills, it's just chat. With skills, it's a real agent.

What ClawX makes easy

You can toggle skills on and off with a single switch, browse the skills marketplace inside the app, open the Skills folder to edit files directly, and install third-party skills cleanly without breaking the agent.

My core skill stack

The skills I always have on are web search for research and citations, browser for site interaction and form filling, memory for cross-session context, and terminal CLI for code and dev tasks.

The skills I add per workflow are Tavly for better web search than the default, file operations for reading and writing local files, calendar for scheduling tasks, and Notion for writing drafts directly into my workspace.

I cover the skills setup deeper in OpenClaw Course.

Adding A New Skill

Go to the Skills tab, click marketplace, search for what you want, and click install.

For non-marketplace skills, open the Skills folder directly from the app, drop in the skill file, and restart ClawX. That's the whole flow.

🔥 Want my full skill stack for ClawX OpenClaw? Inside the AI Profit Boardroom, I share my exact skills config, custom skills I've built, and a 6-hour OpenClaw course covering every skill in detail. Plus weekly screen-share coaching. 2,800+ members. → Get the stack

Channels In ClawX OpenClaw

This is what makes OpenClaw actually useful in production. A chat in the app is fine, but a chat in WhatsApp from your phone is gold.

Channels ClawX supports

The supported channels are WhatsApp, Discord, Telegram, Slack via skill, and custom webhooks for integrating with anything else.

Setting up WhatsApp

The original gateway version of this was painful. In ClawX, you click Channels, click WhatsApp, follow the QR code prompt, and you're done. Took 4 minutes for me first time.

Setting up Discord

Even simpler than WhatsApp. Click Channels, click Discord, paste your bot token, pick the server, and you're live.

What I do with channels

My OpenClaw agent answers WhatsApp messages from my team, Discord pings about content, and Telegram alerts I forward to it for triage. It runs all of this through ClawX in the background.

If you're doing similar with Hermes Open Web UI, the principle is the same — channels are where agents earn their keep.

Models In ClawX OpenClaw

The Models tab is where ClawX really beats the gateway.

Adding providers

In ClawX you click Models, click "Add provider", pick from Anthropic, OpenAI, OpenRouter, Mistral, DeepSeek, Z AI, Ollama, and the rest, paste your API key (or local URL for Ollama), and save.

I have 6 providers configured at any time.

Switching providers per agent

Each agent in ClawX can use a different model, which is one of the platform's biggest wins. I run my main agent on Anthropic, my research agent on DeepSeek for cheap and fast research, my code agent on Qwen 3.6 for better code output, and my QA agent on a small fast cloud model.

In the gateway, this would be a config-file nightmare.

Token usage tracking

ClawX shows token usage per provider over 7 days, 30 days, and all time. Useful for tracking costs across multiple models.

I noticed I was burning more on one provider than expected and switched to a cheaper alternative for that specific agent.

Local + cloud hybrid

You can mix local Ollama models with cloud providers in one ClawX setup, which is useful if you want some agents free and local while others run on top-tier cloud.

I cover the Ollama side in Ollama Hermes — the same principle applies for OpenClaw.

Three Things You Should Set Up On Day One

If you're new to ClawX OpenClaw, get these three set up immediately.

Add at least three model providers so you don't lock yourself to one. The whole point of ClawX is easy switching, and you can't switch if there's only one option.

Connect at least one channel — WhatsApp or Discord both work. This is where the value compounds because the agent moves from chat-only to actually-useful.

Toggle on web search and memory skills because most workflows need both from day one.

Combined Workflow Example

Let me show you what these three features look like working together.

The setup is an agent called "Content Researcher" running on DeepSeek for cost efficiency, with web search, Tavly, memory, and browser skills enabled, connected to WhatsApp as a channel.

The trigger is a WhatsApp message: "Research the keyword 'free AI agent' and draft a brief."

The flow plays out in seven steps. The channel receives the message. The agent reads it. The web search skill kicks in. Tavly fetches deeper sources. The browser skill checks competitor pages. The memory skill stores key findings. The agent drafts the brief and the reply lands in WhatsApp.

I do nothing. It happens in 4 minutes.

That's three features working together — and that's ClawX OpenClaw at its best.

Common Mistakes With These Features

Three mistakes worth avoiding.

Adding skills you don't need creates more confusion, not more capability. Start lean and add skills as actual needs appear.

Connecting too many channels at once is overwhelming. Pick one, master it, then add the second.

Locking yourself to one model defeats the entire purpose of ClawX. Add 3+ providers from day one.

🚀 Want my full ClawX automation system? The AI Profit Boardroom has my exact skills, channel setups, and model rotations. Daily training drops, weekly screen-share coaching, 6-hour OpenClaw course, and 2-hour Hermes course. 2,800+ members. → Join here

FAQ — ClawX OpenClaw Features

How many skills can ClawX run at once?

There's no hard limit. Performance depends on your machine.

Can I write custom skills for ClawX?

Yes — drop them in the Skills folder and restart ClawX.

What's the easiest channel to start with?

WhatsApp is the most reliable and fastest setup.

Can I run ClawX without any model providers?

Technically yes, but the agent won't respond. Add at least one provider.

Does ClawX support local models?

Yes — point it at your local Ollama URL and it works the same as cloud.

Will my skills work if I switch from the gateway?

Yes — ClawX picks up your existing Skills folder automatically.

Can I set different models per agent?

Yes — that's one of ClawX's main wins over the gateway.