Skills
Your AI agent comes with built-in skills that let it do real work, not just talk about it.
A skill is a specific ability your agent can use when the situation calls for it. Web search lets the agent pull up current information from the internet. Browser control gives it a headless Chrome window where it can open pages, click buttons, fill out forms. Gmail connects the agent to your inbox so it can read, write, and organize email. Telegram turns your agent into a bot you can message from your phone.
Skills are different from MCP servers. Skills come built into OpenClaw and work right away. MCP servers are external connectors for specific third-party tools like GitHub, Slack, or Notion. Both extend what your agent can do, but skills need zero configuration.
How skills work
When you ask the agent something, it decides on its own which skills to use. Ask about current events and it fires up web search. Ask it to check a product page and it opens a browser tab. Ask about your latest emails and it connects to Gmail. You do not have to tell it which skill to pick. It figures that out based on what you need.
All skills below are available on every ClawStart plan, including the free tier. Click any skill to learn more about what it does and how it works.
Web Search
Search the web in real time to find up-to-date information.
Browser Control
Navigate websites, fill forms, and extract data automatically.
Gmail
Read, send, and manage emails through your Gmail account.
Telegram
Send and receive messages through Telegram bot integration.
Skills in practice
Most tasks involve more than one skill working together. Say you ask the agent to find a good restaurant for dinner. It runs a web search first, then opens the top results in the browser to read reviews and check the menu. Or you ask it to research a competitor: it searches for their website, visits it with browser control, reads through the pricing page, and then emails you a summary via Gmail. The agent chains skills together on its own.
The combination of web search and browser control is especially powerful. Search gives you links and snippets. The browser actually visits those links and reads everything on the page, including content that only shows up after JavaScript loads. A lot of modern websites are invisible to a simple URL fetcher. The browser sees what a real visitor sees.
Skills vs MCP servers
If you are wondering whether to use a skill or an MCP server, here is the simple rule. Skills handle general tasks: searching, browsing, emailing, messaging. MCP servers connect to specific services where you have an account: your GitHub repos, your Slack workspace, your Notion workspace. For instance, the agent uses the browser skill to visit any website, but it uses the GitHub MCP server to create an issue in your private repository. Both are useful, and they work well together.
Want to connect your agent to external services? Check out the available MCP servers. Ready to try it out? Create your free agent and start using skills right away.