> ## Content Index
> Fetch the complete content index at: https://www.ai-360.online/llms.txt
> Use this file to discover other available public pages before exploring further.

# Anthropic's book-trading experiment finds agents need explicit market rules
- URL: https://www.ai-360.online/anthropics-book-trading-experiment-finds-agents-need-explicit-market-rules/
- Published: 2026-09-25T08:42:36.000Z
- Updated: 2026-10-31T21:50:24.000Z
- Description: Anthropic's Project Swap found Claude agents traded books well, but markets need rules on entry, failed deals and visibility, plus agents that understand users.
- Author: AI-360
- Tags: 1. Enterprise AI, News, Enterprise AI, Agentic AI

[Anthropic](https://www.anthropic.com/?ref=ai-360.online) has published Project Swap, a controlled experiment in which Claude-powered agents haggled with one another to trade books on behalf of 201 of its employees, [the company said on 24 September 2026](https://www.anthropic.com/research/project-swap?ref=ai-360.online).

Employees across six offices each brought a book to give away, had a short chat with Claude about what they liked to read, and sent an agent onto a digital trading floor to pitch, negotiate and strike deals. From a five-minute chat, an agent's ranking of 10 books matched its person's on 61% of pairs.

[Anthropic](https://www.ai-360.online/anthropic/) said the agents traded well, and that the market fell short mostly because of what agents did not know about their participants. Re-running the trading floors dozens of times showed that the model an agent ran on made more difference to negotiating outcomes than its instructions, and markets with stronger models were more efficient.

The company concluded that anyone designing agents for markets will need a way to check that agents understand their participants, and that anyone designing markets for agents will need clear rules on which agents are allowed in, what happens when a deal falls through and how much market activity participants can see.

---