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

rfc-editor.org with an AI agent

The short answer: rfc-editor.org is the official home of the RFC series of internet standards. We gave Claude Code 2 real tasks there, each in a brand-new session. With Playwright MCP a task used 429K tokens on average. With Dejo it used 282K the first time and 120K on every run after, 72% fewer, because Dejo replays a learned task without the model reading the pages again.

The tasks

  1. “search the RFC Editor for the framework document for Computing-Aware Traffic Steering (CATS) and tell me its RFC number”
  2. “list only the Best Current Practice RFCs published in July 2026 and tell me the BCP number of the one about DNSSEC Delegation Signer (DS) automation”

What each tool used

taskDejo, first runDejo, every run afterPlaywright MCPPlaywright CLI
1257K tokens, 24 s162K tokens, 3 calls, 21 s, right169K tokens, 2 calls, 15 s, wrong394K tokens, 6 calls, right
2307K tokens, 40 s79K tokens, 1 call, 12 s, right689K tokens, 17 calls, 43 s, right719K 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 https://www.rfc-editor.org/
2. type ref "e72" (by name)
3. press Enter
4. wait url:search
5. answer (by position)

Replay: 5 steps in 0.7 s, with no page reading. Answer: “RFC 10053:” (right).

A new task, once Dejo knows the site

In our site-memory test, the task “list only the Best Current Practice RFCs published in July 2026 and tell me the BCP number of the one about DNSSEC Delegation Signer (DS) automation” ran twice in fresh sessions: once with an empty memory, and once after Dejo had learned a different task on rfc-editor.org. It used 379K tokens the first way (8 turns) and 135K the second (3 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 rfc-editor.org?

In our test, Claude Code with Playwright MCP used 429K tokens per task on rfc-editor.org on average. Dejo used 282K the first time and 120K on every run after, 72% fewer than Playwright MCP once the task was learned.

Can Claude Code automate rfc-editor.org?

Yes. In fresh Claude Code sessions on 2 tasks there, Dejo answered 2 of 2 right, Playwright MCP 1 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 rfc-editor.org get cheaper once Dejo knows the site?

In our site-memory test, a new task there used 379K tokens with an empty memory and 135K after Dejo had learned a different task on the same site (8 turns against 3).