Agents find you, but often give up before they finish a task.
You · 62.2
InvisibleWeakPartialStrongAgent-ready
Where your 62.2 points come from
Can agents find, reach and trust your site?
Agents searching for your docs don’t find themFailedMust-have+2.9est. points
Scan result
None of the top 10 results for an API docs search is your own documentation.
What this checks
Do the searches an agent would actually run lead to your developer resources? The scan looks up your API docs and expects one of the results to be your own docs page, then does the same for your SDK, CLI or MCP server and, if one exists, ex
Agents can’t see everything you offer in one placeFailedMust-have+1est. points
Scan result
There is no agent resource catalog: /.well-known/ard.json, a homepage rel="ard" link and the legacy /.well-known/ai-catalog.json are all absent.
What this checks
A single file listing everything agents can use from you: MCP servers, APIs, agents and skills. Lacking one, an agent has to piece your capabilities together page by page. The scan follows ARD v0.91: a catalog at /.well-known/ard.json, or o
Searching your name doesn’t lead agents to youWarningMust-have+1est. points
Scan result
specsight.app sits at position 6 for "Specsight", below the leading results.
What this checks
A search for your brand name plus your main product should return your domain near the top. When it doesn't, agents can't dependably distinguish you from lookalikes and resellers.
Developers’ agents can’t install your SDKFailedMust-have+1est. points
Scan result
No SDK or CLI package for your product turned up on npm or PyPI.
What this checks
A first-party SDK or CLI on PyPI or NPM hands agents, along with the developers directing them, a prebuilt entry point rather than writing each API call by hand.
Agents can’t confirm who you are from a trusted sourceFailedNice-to-have+0.9est. points
Scan result
No Wikipedia article or Wikidata item for "specsight" links to specsight.app.
What this checks
A Wikipedia article plus a Wikidata item, both linking to your domain. Wikipedia supplies more citations to AI answers than any other single source (~48% of ChatGPT citations), meaning a presence there decides whether an AI can confirm who
Your site can’t tell trusted agents from botsFailedEarly standard+0.2est. points
Scan result
/.well-known/http-message-signatures-directory is not published.
What this checks
A published directory of public keys (RFC 9421) that agents use to sign their requests cryptographically, letting you identify each caller precisely and admit the legitimate ones.
Agents can’t complete the sign-up flow×Failedagent-onboardingMust-haveNo machine-readable data was found for this check on the homepage
+2.1est. points
Agents get partial answers from your docs!Warningdocs-coverageNice-to-haveNo machine-readable data was found for this check on the homepage
+1.4est. points
Agents can’t verify your pricing tiers×Failedpricing-schemaMust-haveNo machine-readable data was found for this check on the homepage
