Agent Terms File

Simulated: the sites, the CDN, the agents, the signatures, and the payments are all made up. Nothing here talks to a real service.

1. Pick a site

2. Send an agent to the site

Request path: agent, CDN, site An agent asks for a page. The request goes to the CDN. The CDN reads the terms file of the site and the identity of the agent. It then allows the request, asks for payment, or denies it. Agent CDN Site Ready

    Terms JSON and verdict

    
                  
                  
    
                

    3. Your rules

    4. Result

      Cookie banners were built for people. Agents handle them inconsistently. A terms file gives the CDN and the browser one place to read the rules.

      Try this

      1. Pick "Recipe blog" and "Perplexity". Choose "Unsigned: it sends its real name". Press Play. The CDN denies it.
      2. Choose "Unsigned: it claims to be Google". Press Play. The lie works, because nothing proves the name.
      3. Pick "Ticket shop" and "OpenAI", signed. Press Play. Read the 402 notice, then approve the payment.
      4. Press "Check the terms" to compare the file with your rules.

      Pain point

      A publisher that blocks AI crawlers has no clean way to say yes for a price.

      Problem statement

      A site cannot say which AI agents may read it, on what terms, and at what price. The network has no terms to enforce.

      What sparked it

      HTTP already defines the status code 402 Payment Required. An IETF working group (webbotauth) works on signed agent requests.

      Why this approach

      The site publishes one terms file. The CDN reads it at the edge, checks who the agent is, and answers.

      Hypothesis tested

      Assumption: we believed that one terms file lets a CDN answer allow, payment required, or deny. If it holds, a publisher can charge for agent reads without a custom setup per agent.

      How to try it

      1. Open the Prototype tab. Pick a site, an agent, a CDN, and an identity.
      2. Press Play, or press Step.
      3. Read the chat and the "Terms JSON and verdict" panel.
      4. Press "Check the terms" to compare the file with your rules.

      What this does not do

      Notes

      Traces

      Prompts used to make this prototype
      StepPromptModelToolDate

      Cost

      Cost to make this prototype
      StepModelInput tokensOutput tokensCost (USD)
      Total

      Measured: the author runs used 194,609 tokens in total. Estimated: the split into input and output tokens (90 percent input, 10 percent output) and the dollar cost, at $2 input and $10 output per million tokens for Sonnet 5.5 (about $0.54). The run reported one total only. The lead session's planning, research, and publishing are not included. The prompts in the Traces section are listed in the order they were given.