Step 01: Give your agent a tool.

Simple tools are more powerful than you think. Read, Write, Bash is enough.

Prerequisites

Same as Step 00 - copy the config file and add your API key:

cp default_workspace/config.example.yaml default_workspace/config.user.yaml
# Edit config.user.yaml to add your API key

What We will Build?

Giving the agent the ability to actually do things, from chatting only to taking real actions.

Key Components

Stop Reason: Chat loop branches on stop_reason — "tool_calls" to execute tools, "stop" for normal completion, "length" for truncated responses
Tools: Manages available tools and executes tool calls
Tool Calling Loop: Agent calls tools, adds results to history, continues conversation

src/mybot/tools/base.py

class BaseTool(ABC):
    name: str
    description: str
    parameters: dict[str, Any]

    @abstractmethod
    async def execute(self, session: "AgentSession", **kwargs: Any) -> str:
        pass

    def get_tool_schema(self) -> dict[str, Any]:
        return {
            "type": "function",
            "function": {
                "name": self.name,
                "description": self.description,
                "parameters": self.parameters,
            },
        }

src/mybot/core/agent.py

Integrating Tools into Chat Loop.

class AgentSession:
    async def chat(self, message: str) -> str:
        user_msg: Message = {"role": "user", "content": message}
        self.state.add_message(user_msg)

        tool_schemas = self.tools.get_tool_schemas()
        logger = logging.getLogger(__name__)

        while True:
            messages = self.state.build_messages()
            content, tool_calls, stop_reason = await self.agent.llm.chat(messages, tool_schemas)

            assistant_msg: Message = {
                "role": "assistant",
                "content": content,
                "tool_calls": [...],
            }
            self.state.add_message(assistant_msg)

            if stop_reason == "tool_calls":
                await self._handle_tool_calls(tool_calls)
                continue

            if stop_reason == "length":
                logger.warning(
                    "LLM response truncated (max_tokens reached), "
                    "returning partial response"
                )

            break

        return content

Notes

Why Tools Are Hardcoded

Tools are registered via ToolRegistry.with_builtins() in Python, not declared in AGENT.md frontmatter.

YAML-based tool config involves design decisions beyond the scope of this tutorial. The BaseTool abstraction already keeps tools pluggable in code.

OpenClaw handles this with a plugin system and per-agent JSON config — but that's out of scope for this tutorial.

Try it out

cd 01-tools
uv run my-bot chat

# You: Hey Can you read your README.md please?
# pickle: I found and read the README.md file! 🐱

# # Step 01: Tools - Read, Write, Bash is Powerful Enough

# Give the agent the ability to execute tools (read, write, edit, bash) and interact with the filesystem.
# [More lines]

What's Next

Step 02: Skills - Add dynamic capability loading with SKILL.md files.