22 of 42
ContentsPart III · Chapter 10

Chapter 10Tools

The agent loop

The model never touches anything. It asks the harness for a tool, the harness decides and does it, and the result goes back in as more text.

The agent loop, step by step

Illustration: message shapes real; the scenario is made up

Illustration. The message shapes are trimmed from each company's documentation;12 the file, the numbers and the web page are made up.

Go deeper
  • MCP was introduced in November 20246 and donated in December 2025 to the Agentic AI Foundation, which Anthropic describes as a directed fund under the Linux Foundation, co-founded by Anthropic, Block and OpenAI.7
  • At that point, Anthropic said, its official SDKs had over 97 million monthly downloads.7
  • The idea of a model reasoning, acting with a tool, observing and repeating was formalized in 2022 research called ReAct.8
Notes (8)
  1. 1Anthropic, “Handle tool calls,” Claude Platform Docs, https://platform.claude.com/docs/en/agents-and-tools/tool-use/handle-tool-calls (accessed October 26, 2026).
  2. 2OpenAI, “Function calling,” OpenAI API docs, https://developers.openai.com/api/docs/guides/function-calling (accessed October 26, 2026).
  3. 3Kai Greshake, Sahar Abdelnabi, Shailesh Mishra, Christoph Endres, Thorsten Holz, and Mario Fritz, “Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection,” in Proceedings of the 16th ACM Workshop on Artificial Intelligence and Security (November 30, 2023), 79–90, https://doi.org/10.1145/3605764.3623985.
  4. 4Anthropic, “Use Claude Cowork safely,” Claude Help Center, https://support.claude.com/en/articles/13364135-use-claude-cowork-safely (accessed October 26, 2026).
  5. 5OpenAI, “OpenAI Model Spec (2025-10-27 version),” October 27, 2025, https://model-spec.openai.com/2025-10-27 (accessed October 26, 2026).
  6. 6Anthropic, “Introducing the Model Context Protocol,” November 25, 2024, https://www.anthropic.com/news/model-context-protocol (accessed October 26, 2026).
  7. 7Anthropic, “Donating the Model Context Protocol and establishing the Agentic AI Foundation,” December 9, 2025, https://www.anthropic.com/news/donating-the-model-context-protocol-and-establishing-of-the-agentic-ai-foundation (accessed October 26, 2026).
  8. 8Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao, “ReAct: Synergizing Reasoning and Acting in Language Models” (preprint, arXiv, October 6, 2022), https://doi.org/10.48550/arXiv.2210.03629.

Chapter 10 · page 2 of 2

Turn pages with the ← and → keys.

Cite this page

Using this book

© 2026 Steve Sparks. All rights reserved. You're welcome to quote short passages and cite any page, with a link back. To reuse diagrams or larger parts, ask through GitHub.

Quoted material and figures from other sources remain their owners'.