Why tools change everything

Imagine you have a very intelligent assistant locked in a room with no phone, no computer, nothing. They can think brilliantly, but can’t do anything. You ask “find me cheap flights to Lisbon” and they say “great idea, but I don’t have internet access”.

That’s an LLM without tools. Intelligent but useless for practical tasks. Tools are the agent’s phone and computer.

What is a tool?

A tool is a specific function the agent can execute. Technically, it’s a piece of code that does something concrete and that the model can “call” when it needs to.

Common examples: Web search. Read URLs. Send Telegram messages. Read/write files. Execute commands.

What is a skill?

In the OpenClaw context, a skill is the name given to a package of related tools. For example, a “Reddit” skill might include: searching posts, reading comments, getting popular posts.

How does the agent decide which tool to use?

This is one of the most impressive capabilities of modern LLMs. The model doesn’t follow rigid rules — it reasons about which tool it needs.

You send: “What’s in the news about AI today?” → The model “thinks”: I need current information, I have a web search tool, I’ll use it. And it performs the search.

It can even chain tools: search for information, save it to a file, and send you a message to let you know.

MCP: the standard that connects everything

MCP (Model Context Protocol) is a standard created by Anthropic that defines how agents connect to external services.

Before MCP, every integration was a custom project. MCP standardises the communication. A service implementing MCP can connect to any agent supporting MCP — no custom code needed. Think of MCP as the USB of AI tools: a universal connector.

Today there are MCP servers for hundreds of services: Google Calendar, Gmail, Slack, GitHub, databases, and many more.

Where to find tools

Clawhub.ai: The skill directory for OpenClaw. Community MCP servers. Custom skills if you have technical knowledge.

Precautions

Principle of least privilege: Only give the agent the tools it actually needs.

Review what it does: Especially at first, monitor how the agent uses tools.

Sensitive tools: Tools that send emails or publish content can cause problems if the agent misuses them.

Key concepts today

  • Tool: A specific function the agent can execute
  • Skill: A package of related tools (OpenClaw terminology)
  • MCP (Model Context Protocol): Standard for connecting agents to external services
  • The agent decides which tool to use — it doesn’t follow rigid rules, it reasons about the best option
  • Principle of least privilege: Only give it the tools it needs

Next lesson: An agent’s memory — how it remembers (and forgets).