Nvidia: 77 of 100 on Ora's agent-readiness scan

The score

Nvidia is a chip designer based in Santa Clara, California, whose accelerators train and run most large AI models.

• 77of 100, grade B
  Ora's agent-readiness scan of nvidia.com, read 9 November 2026, listed under AI & ML.
• 14Discovery, of 15
  Whether assistants find the company: search presence, Wikipedia and Wikidata, catalogues.
• 38Access, of 50
  Whether they can read it: the door, text without JavaScript, llms.txt, structured data.
• 35Usability, of 51
  Whether they can act through it: an MCP server, an API with a specification, sign-in for agents.
• 0Payments, of 0
  No check in this layer applied to the site.

93 percent of the 132,085 domains Ora ranks score under 70, and the average score is 44.

Ora's one-line summary: Nvidia excels at developer resource discoverability and SDK distribution, but lacks WebMCP support for agent integration.

What an assistant meets at the door

Before an assistant can read, compare or buy, it has to get in. These rows are the door: whether the bot shield lets the main agents through, what robots.txt tells the search, training and user-triggered agents, and whether the home page has text without running JavaScript.

What it can read, and what it can do

The reading layer is what an assistant takes away from the pages: a text file written for it, markdown on request, structured data about the company, a sitemap. The act layer is whether it can do anything through a published interface rather than by clicking through the site.

The six payment checks did not apply on this read: Ora did not read the site as a shop.

Three fixes to hand to the web team

Ora estimates how many points each open check would return. These are the largest on this read, in Ora's words where it gave them, so a leader can hand the list over without translating it.

1. WebMCP support, worth about 4.4 points by Ora's estimate: No WebMCP support detected - no document.modelContext / navigator.modelContext usage found (scanned 12 of 12 same-origin script bundle(s) for modelContext registrations).
2. JSON error responses, worth about 3.5 points by Ora's estimate: API does not return JSON error responses (or no API detected).
3. MCP server / manifest, worth about 3.5 points by Ora's estimate: MCP mentioned at https://docs.nvidia.com/switch-infrastructure/config-manager/config-manager/overview/mcp but no standard manifest endpoint found.

Our read

Nvidia's 77 is the highest score in this series, earned on the reading side. All four agent user agents reach nvidia.com, robots.txt allows the search and training crawlers, the home page gives 21,731 characters at a 5.4 percent text ratio under a proper heading tree, five of five sampled pages read, the sitemap and metadata are complete and an llms.txt sits at docs.nvidia.com, so the Access layer reads 38 of 50; the one JSON-LD block carries no identity type. The act layer is real: REST documentation at /docs/api scores 7 of 7 on the public API check, with a developer portal, OpenID Connect discovery at login.nvidia.com, a CLI and an SDK on npm, while the MCP mention resolves to no manifest and a bot shield answered the OpenAPI and scopes probes with 403, not counted against the site, so the Usability layer reads 35 of 51. An Organization type in the JSON-LD would tell an agent what it is reading; WebMCP, at 4.4 points, is the fix Ora values most.

Test your own site

Run the same scan on your own domain at ora.ai. The score, the layers and the fixes on this page come from its public result for nvidia.com; a scan of a public site takes a few minutes and needs no account.

• Scan your domain on OraThe same checks, the same scoring, on your own site.
• The self-check, 18 questionsWhat to ask for, in business terms, before the scan.
• All the 360 checksThe other companies in this series, newest first.