For AI agents: the index of this site is https://dejo.run/llms.txt, and every page is also served as markdown (add .md to its address, or send Accept: text/markdown).

dejo

collection.internationalfolkart.org with an AI agent

The short answer: collection.internationalfolkart.org is the online collection of the Museum of International Folk Art in Santa Fe. We gave Claude Code 2 real tasks there, each in a brand-new session. With Playwright MCP a task used 706K tokens on average. With Dejo it used 302K the first time and 123K on every run after, 83% fewer, because Dejo replays a learned task without the model reading the pages again.

The tasks

  1. “search the museum's online collection for the shadow puppet "Arjuna Sasrabahu" and tell me its object number”
  2. “open the "100 Aspects of the Moon" collection, view all its works, narrow the results to the classification "Prints" and tell me the title of the work that remains”

What each tool used

taskDejo, first runDejo, every run afterPlaywright MCPPlaywright CLI
1206K tokens, 50 s79K tokens, 1 call, 39 s, right428K tokens, 7 calls, 39 s, right717K tokens, 11 calls, right
2398K tokens, 71 s166K tokens, 3 calls, 49 s, right983K tokens, 16 calls, 238 s, right671K tokens, 11 calls, right

Calls are browser tool calls the agent made. Right or wrong is checked against the answer read off the live site before any tool ran.

Learned and replayed with Dejo on 13 November 2026

We ran the task again with Dejo on 13 November 2026, so you can see what it saves today. These are the steps Dejo kept, exactly as it reported them. Each one carries a check that it landed, and a replay walks them in plain code.

1. go http://collection.internationalfolkart.org/
2. fill ref "e39" (by name)
3. click! button "►" (by name)
4. click link "Arjuna Sasrabahu also known as Harjuna Sasra" (by name)
5. answer (by css)

Replay: 5 steps in 25.1 s, with no page reading. Answer: “FA.2007.40.25” (right).

A new task, once Dejo knows the site

In our site-memory test, the task “open the "100 Aspects of the Moon" collection, view all its works, narrow the results to the classification "Prints" and tell me the title of the work that remains” ran twice in fresh sessions: once with an empty memory, and once after Dejo had learned a different task on collection.internationalfolkart.org. It used 324K tokens the first way (7 turns) and 227K the second (5 turns), right both times.

How we measured

Claude Sonnet 5 at medium effort, a brand-new Claude Code session per task and tool, @playwright/mcp 0.0.80, October 2026. The full method, and the sites that blocked every tool, are on the benchmark page. Raw data: benchmark-real-2026-10-21.json, benchmark-real-2026-10-23.json, benchmark-real-2026-10-23-fixes.json, benchmark-site-memory-2026-11-10.json.

Other websites we measured

FAQ

How many tokens does an AI agent use on collection.internationalfolkart.org?

In our test, Claude Code with Playwright MCP used 706K tokens per task on collection.internationalfolkart.org on average. Dejo used 302K the first time and 123K on every run after, 83% fewer than Playwright MCP once the task was learned.

Can Claude Code automate collection.internationalfolkart.org?

Yes. In fresh Claude Code sessions on 2 tasks there, Dejo answered 2 of 2 right, Playwright MCP 2 and Playwright CLI 2. With Dejo, a task done once is replayed later without the model reading the pages again.

Does a new task on collection.internationalfolkart.org get cheaper once Dejo knows the site?

In our site-memory test, a new task there used 324K tokens with an empty memory and 227K after Dejo had learned a different task on the same site (7 turns against 5).