The Agent Loop

[ s01 ] s02 > s03 > s04 > s05 > s06 | s07 > s08 > s09 > s10 > s11 > s12

"One loop & Bash is all you need" -- one tool + one loop = an agent.

Harness layer: The loop -- the model's first connection to the real world.

Problem

A language model can reason about code, but it can't touch the real world -- can't read files, run tests, or check errors. Without a loop, every tool call requires you to manually copy-paste results back. You become the loop.

Solution

+--------+      +-------+      +---------+
|  User  | ---> |  LLM  | ---> |  Tool   |
| prompt |      |       |      | execute |
+--------+      +---+---+      +----+----+
                    ^                |
                    |   tool_result  |
                    +----------------+
                    (loop until stop_reason != "tool_use")

One exit condition controls the entire flow. The loop runs until the model stops calling tools.

How It Works

1. User prompt becomes the first message.

messages.push({ role: "user", content: query });

1. Send messages + tool definitions to the LLM.

const response = await client.messages.create({
  model: MODEL, system: SYSTEM, messages,
  tools: TOOLS, max_tokens: 8000,
});

1. Append the assistant response. Check stop_reason -- if the model didn't call a tool, we're done.

messages.push({ role: "assistant", content: response.content });
if (response.stop_reason !== "tool_use") {
  return;
}

1. Execute each tool call, collect results, append as a user message. Loop back to step 2.

const results: Anthropic.Messages.ToolResultBlockParam[] = [];
for (const block of response.content) {
  if (block.type === "tool_use") {
    const input = block.input as { command: string };
    const output = runBash(input.command);
    results.push({
      type: "tool_result",
      tool_use_id: block.id,
      content: output,
    });
  }
}
messages.push({ role: "user", content: results });

Assembled into one function:

async function agentLoop(messages: Anthropic.Messages.MessageParam[]): Promise<void> {
  while (true) {
    const response = await client.messages.create({
      model: MODEL, system: SYSTEM, messages,
      tools: TOOLS, max_tokens: 8000,
    });
    messages.push({ role: "assistant", content: response.content });

    if (response.stop_reason !== "tool_use") {
      return;
    }

    const results: Anthropic.Messages.ToolResultBlockParam[] = [];
    for (const block of response.content) {
      if (block.type === "tool_use") {
        const input = block.input as { command: string };
        const output = runBash(input.command);
        results.push({
          type: "tool_result",
          tool_use_id: block.id,
          content: output,
        });
      }
    }
    messages.push({ role: "user", content: results });
  }
}

That's the entire agent in under 30 lines. Everything else in this course layers on top -- without changing the loop.

What Changed

Component	Before	After
Agent loop	(none)	while (true) + stop_reason
Tools	(none)	bash (one tool)
Messages	(none)	Accumulating list
Control flow	(none)	stop_reason !== "tool_use"

Try It

cd learn-claude-code-ts
bun run agents/s01_agent_loop.ts

1. Create a file called hello.ts that prints "Hello, World!"
2. List all TypeScript files in this directory
3. What is the current git branch?
4. Create a directory called test_output and write 3 files in it