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

catalog.ccv.edu with an AI agent

The short answer: catalog.ccv.edu is the course catalog of the Community College of Vermont. We gave Claude Code 2 real tasks there, each in a brand-new session. With Playwright MCP a task used 1.2M tokens on average. With Dejo it used 402K the first time and 83K on every run after, 93% fewer, because Dejo replays a learned task without the model reading the pages again.

The tasks

  1. “in the 2026-2027 course list, show only the MEC courses, open MEC 2350 Green Production and tell me how many credits it is worth”
  2. “open the full course list in the 2026-2027 catalog, go to page 3 and tell me the last course listed on that page”

What each tool used

taskDejo, first runDejo, every run afterPlaywright MCPPlaywright CLI
1324K tokens, 38 s83K tokens, 1 call, 19 s, right1.4M tokens, 20 calls, 72 s, right1.2M tokens, 16 calls, right
2480K tokens, 44 s82K tokens, 1 call, 23 s, right1.0M tokens, 15 calls, 66 s, right973K tokens, 13 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 https://catalog.ccv.edu/
2. click link "Courses" (by name)
3. select combobox "Choose Course Prefix" "MEC" (by name)
4. click button "Filter" (by name)
5. click link "MEC 2350 - Green Production" (by name)
6. wait text:Credits:
7. answer (by position)

Replay: 7 steps in 8.1 s, with no page reading. Answer: “MEC 2350 - Green Production Credits: 1 This course provides an overview of environmentally…” (right).

A new task, once Dejo knows the site

In our site-memory test, the task “open the full course list in the 2026-2027 catalog, go to page 3 and tell me the last course listed on that page” ran twice in fresh sessions: once with an empty memory, and once after Dejo had learned a different task on catalog.ccv.edu. It used 414K tokens the first way (8 turns) and 301K the second (6 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 catalog.ccv.edu?

In our test, Claude Code with Playwright MCP used 1.2M tokens per task on catalog.ccv.edu on average. Dejo used 402K the first time and 83K on every run after, 93% fewer than Playwright MCP once the task was learned.

Can Claude Code automate catalog.ccv.edu?

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 catalog.ccv.edu get cheaper once Dejo knows the site?

In our site-memory test, a new task there used 414K tokens with an empty memory and 301K after Dejo had learned a different task on the same site (8 turns against 6).