Agent readiness · Commerce platforms
fourthwall.com
11 of 13 measurable points under formula 9.57, measured on 2026-09-03. 16 checks exist; each is one HTTP request with a published rule, so every sentence below can be rerun and argued with. 3 of them could not be measured from where we ask, and those are left out of the denominator rather than counted as failures.
A · Discovery
- PASSAnswers an agent user-agentAnswered 200 to LetAgentsIn/1.0 (+https://letagentsin.com/bot)
- FAILPrice or entry condition in the search snippetYour pricing page's description tag names no amount and no entry condition: “No monthly fees. No upfront costs. We make money when you do.”Put the number, or the entry condition in words ("free tier", "no credit card"), into the pricing page's meta description. That string is what an agent reads when it decides whether to open you at all.
- PASSllms.txt publishedllms.txt and llms-full.txt present at https://fourthwall.com/llms.txt and https://fourthwall.com/llms-full.txt and https://docs.fourthwall.com/llms.txt and https://docs.fourthwall.com/llms-full.txt, and the 12 links we sampled across the 4 files all answer
- PASSDocs readable without JavaScript8,067 characters of text without JS, on https://docs.fourthwall.com/api-reference/platform/shop/get-public-token rather than on https://docs.fourthwall.com
- PASSOn-demand agents not blockedNo on-demand agent is blocked
- N/APaths robots.txt points at answerrobots.txt names no concrete path, only patterns or nothing, so there is no claim to check
- PASSNo punishing crawl delayNo Crawl-delay applies to the agents we check
B · Agent entry
- PASSAgent entry pointFound: https://docs.fourthwall.com/skill.md
- PASSOAuth dynamic client registrationregistration_endpoint published, but none of the 2 advertised grants (authorization_code, refresh_token) finishes without a person at a browserAdvertise client_credentials on the same metadata for the clients that should run unattended, and an agent that registered itself can reach a token without a person.
- PASSMCP surfaceLive MCP endpoint at https://mcp.fourthwall.com, answered 401 with an auth challenge
C · Registration
- UNMEASUREDNo CAPTCHA in the signup HTMLUnmeasurable: the signup form at https://fourthwall.com/get-started is not in the server HTML, so its gates are not eitherServer-render the form, or tell us the endpoint it posts to, and the gates become visible to us and to an agent.
- FAILSignup reachable without a browserhttps://fourthwall.com/get-started is reachable, but its form needs JavaScript
D · Provisioning
- UNMEASUREDProgrammatic key provisioningUnmeasurable: none of the 4 documentation pages we reached is about keys or authentication, so their silence about creating one says nothingLink your API key or authentication page from your docs index or from llms.txt and this becomes measurable.
- PASSFree tier or no-card trial stated in textFree tier or no-card signals at https://fourthwall.com/pricing: "Free forever", "$0"
E · Integration
- UNMEASUREDTyped SDK on the registryUnmeasurable: we could not identify the package a developer installs to use you, from your site, your docs or a registry searchName your package once in your docs, or link it from your repository, and we stop guessing. If your product is not something a developer installs, this check does not apply to you: unmeasured is out of the score and out of the denominator, so it costs you nothing.
- PASSMachine-readable API descriptionDocs serve markdown to machines, at https://docs.fourthwall.com
Tell me when this changes
The failures here are the kind nobody notices. An edge rule that starts refusing agents changes nothing a person sees in a browser, so the first sign is usually an integration that quietly stopped working. We rescan weekly and write only when a verdict moves.
Free while we are building this, and we will ask before it ever costs anything. No card, no account, and one link in every email that stops it.
This page is the newest scan we hold for fourthwall.com and changes when we rescan. It is not a judgement of the product: we measure whether an unattended run can get through, not whether the thing is any good. The whole corpus is published as JSON and CSV.
If a line here is wrong, rescan it first, because most of what we get told is already fixed and the page is only as new as the last scan. If it is still wrong after that, write to hello@letagentsin.com with the check and what you see instead. A verdict we cannot reproduce comes down, and the rule that produced it gets fixed for everybody rather than only for you.