--- title: 100/100 on ora.ai: what the last six points cost description: On 10 September 2026 ora.ai scored watchfor.io 100/100 for agent-readiness and listed it first among the 90,000-odd domains it has scanned. We are proud of that. This is the honest version: what the last six points took, why we sit in the Community category, and what a 100 does not mean. canonical: https://watchfor.io/blog/100-on-ora-ai --- # 100/100 on ora.ai: what the last six points cost 2026-09-11 · WatchFor Team · Engineering · engineering, architecture, monitoring On 10 September 2026 ora.ai scored watchfor.io 100/100 for agent-readiness and listed it first among the 90,000-odd domains it has scanned. We are proud of that. This is the honest version: what the last six points took, why we sit in the Community category, and what a 100 does not mean. On 10 September 2026, [ora.ai](https://ora.ai/score/watchfor.io) — an independent index that grades how discoverable and usable a domain is for AI agents — scored watchfor.io **100/100, A+**, and put it at the top of its leaderboard. A day later it still does, and the report page reads "#1 of 90,401". We are a small team, and we are proud of that. We are also engineers, so the rest of this post is the part a press release would leave out: what the last six points actually cost, and exactly what the number does and does not say. ## Three things to read before the number **We are listed in "Community", not in a named category.** Every scanned domain starts there. A named category — Infrastructure, Developer Tools — requires a [Tranco](https://tranco-list.eu/) top-50,000 traffic rank and a verification review, and we do not have that traffic yet. The consequence is concrete: the leaderboard's default view hides Community listings, so you have to tick *Include community* to see us at all. In the default view the top score belongs to Telnyx. **Ties at 100 reorder.** On 10 September six domains held 100 and we were listed first; by 11 September the set and the order had already changed. ora does not publish its tiebreak. So the accurate claim is "one of a handful at 100, listed first on the day" — not "#1", full stop. **The rubric moves under you.** In August we wrote up [the week that took us from 40 to an A](/blog/what-it-took-to-make-monitoring-agent-ready). We reached 94 on 29 August and 97 on 2 September — and then 95 on 4 September without changing a line, because ora added two checks. A score is a reading against the rubric of that day, which is why every number in this post carries one. ## The last six points 94 to 100 took twelve days and looked nothing like the first fifty. **Our own CDN was silently failing the scanner.** Rescans kept returning a stale snapshot. The cause was a browser-integrity check at the edge that answered 403 to non-browser user agents — invisible from a browser, fatal to a crawler. Turning it off was one toggle, and the hardest of the six points to find, because nothing on our side logged it. **Path parameters that agents could not see.** ora converts an OpenAPI document into function-calling definitions and, it turns out, reads only operation-level parameters. Our `{id}` parameters were declared at the path level — valid OpenAPI, and every detail endpoint looked argument-less to an agent. A post-processing step now hoists them, and a test guards it. **Our own brand name was missing.** Every marketing page overrode its Open Graph metadata without `siteName`, and Next.js does not inherit it, so scanners and link previews called us "watchfor.io" rather than "WatchFor". One root cause, 53 pages. **Two new checks, both about search.** The September drop came from *agentic-search-specific* and *brand-search-accuracy*: can an agent find our API docs, OpenAPI and auth docs by name? Google could not. We built a [developer portal](/developers), gave every API doc page a search-facing title, a canonical URL and TechArticle structured data, and asked Search Console to re-index. The rest was off-code — a company page here, an org description there. **WebMCP, from 61 to 99.** ora runs a separate audit of in-page agent tools. Ours scored 61, with a Trust score of zero: none of the read-only tools said they were. Version two added `readOnlyHint`, titles, verb-first names, and descriptions that contain no instructions to the agent — an instruction in a tool description is an injection surface. Then the native API turned out to sit behind a Chrome origin trial: without the token `document.modelContext` does not exist and our tools registered only under a shim. With the token, verdict *active*, 10 tools, [99/100](https://ora.ai/api/webmcp/watchfor.io). **A second opinion.** [is-agentic.com](https://is-agentic.com/scan/watchfor.io), a different scanner with a different rubric, scores us 100/100 — 12 of 12 essential checks and 22 of 23 recommended. ## What ora still lists The report is not clean, and it should not read as if it were. ora's own summary line names the gap: no Wikipedia article or Wikidata entity for "WatchFor". Two further warnings we believe sit on the scanner's side — a "thin" `pricing.md` that is 384 lines long, and a tool-count drift that comes from reading our documentation MCP server as the product one — and we have reported both. If they turn out to be ours, we will fix them and say so here. ## Why it matters, in one sentence Not the leaderboard. An agent that has never heard of WatchFor can now find the API, read the docs, obtain a scoped key and run its first authenticated call with no person in the loop — and on 10 September 2026 an independent scanner agreed. --- *ora.ai is independent of us and we have no commercial relationship with it. The score, category and rank above are as displayed on 10–11 September 2026 and will change; the live report is [ora.ai/score/watchfor.io](https://ora.ai/score/watchfor.io).* --- Canonical page: https://watchfor.io/blog/100-on-ora-ai · All articles: https://watchfor.io/blog/all.md · Site guide: https://watchfor.io/llms.txt