<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>The Agent Loop with Claude on DATATWEETS</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/</link><description>Recent content in The Agent Loop with Claude on DATATWEETS</description><generator>Hugo</generator><language>en</language><copyright>Copyright (c) 2025 Datatweets</copyright><lastBuildDate>Sun, 28 Jun 2026 09:00:00 +0200</lastBuildDate><atom:link href="https://datatweets.com/courses/ai-agents/the-agent-loop/index.xml" rel="self" type="application/rss+xml"/><item><title>Lesson 1 - Tool Use with Claude</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-1-tool-use-with-claude/</link><pubDate>Fri, 26 Dec 2025 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-1-tool-use-with-claude/</guid><description>Tool use is the mechanism that turns a chatbot into an agent. Learn to define a tool with a name, description, and JSON-schema inputs, pass it to messages.create, and read the tool_use block Claude returns when it wants to act.</description></item><item><title>Lesson 2 - Handling tool_use and tool_results</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-2-handling-tool-use-and-tool-results/</link><pubDate>Fri, 26 Dec 2025 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-2-handling-tool-use-and-tool-results/</guid><description>Close the loop on a single tool call: detect the tool_use block, run the real function, build a tool_result with a matching tool_use_id, append the assistant turn and then a user turn carrying the result, and call the model again to get the final answer.</description></item><item><title>Lesson 3 - Building the Agent Loop</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-3-building-the-agent-loop/</link><pubDate>Fri, 26 Dec 2025 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-3-building-the-agent-loop/</guid><description>Generalize a single tool call into the agent loop: a function that calls Claude and, while the stop reason is tool_use, runs the requested tools and feeds results back until a final answer — with a max-steps guard. See it run end to end.</description></item><item><title>Lesson 4 - Errors, Parallel Tools, and Stopping Conditions</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-4-errors-parallel-tools-and-stopping/</link><pubDate>Fri, 26 Dec 2025 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-4-errors-parallel-tools-and-stopping/</guid><description>A real agent loop needs guardrails: catch tool errors and return them as is_error results so Claude can recover, run multiple tool calls and return all results in one message, and bound the loop with a max-steps cap so it can never run forever.</description></item><item><title>Lesson 5 - Guided Project: Atlas's First Tool Loop</title><link>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-5-guided-project-atlas-first-tool-loop/</link><pubDate>Fri, 26 Dec 2025 09:00:00 +0200</pubDate><guid>https://datatweets.com/courses/ai-agents/the-agent-loop/lesson-5-guided-project-atlas-first-tool-loop/</guid><description>Put Module 2 together: give Atlas two real tools and the agent loop, wire it to the Anthropic client, and run it on a question that needs both tools — the skeleton every later module extends.</description></item></channel></rss>