OpenAI: the 360 agent-readiness check
OpenAI scores 60 of 100 on Ora's agent-readiness scan, a C, read 9 November 2026. Of the five surfaces an agent acts on, Ora found 3: a public API, OAuth, a command line. A challenge page in front of the company whose crawler every site names.
On 9 November 2026 the public site at openai.com scored 60 of 100 on Ora's agent-readiness scan, a C, in Ora's AI & ML category. Ora probes a site the way an agent would: it requests pages with the agents' own user agents, reads robots.txt, looks for the files agents read, and tries the API, MCP and payment surfaces. This check reads the act layer first, because that is what the directory records, then the door, what an agent can read, and whether the company is found and recommended. Ora's one-line summary: OpenAI offers strong developer SDK and discovery presence, but lacks WebMCP support and MCP Apps UI resources for agents.
The score
| Layer | Asks | Points | Share |
|---|---|---|---|
| Usability | Can agents use it | 31 of 62 | 50% |
| Access | Can agents read it | 29 of 54 | 54% |
| Discovery | Can agents find and recommend it | 14 of 16 | 88% |
| Payments | Can agents pay | 0 of 0 | none apply |
| Total | Grade C | 60 of 100 | 60% |
What an agent can do
The surfaces an agent acts on: an API it can find and call, an MCP server, an agent card, OAuth metadata, a command line. This is the layer the directory records, so it comes first.
In the directory, OpenAI no MCP endpoint on record; the connector run of 4 November 2026 is on the entry page. Ora's MCP server check for the same domain reads partial; the two look for different things, the directory for a server that answers a handshake and lists tools, Ora for a server it can find from the site.
| Check | Result | Points | What Ora saw |
|---|---|---|---|
| MCP server / manifest mcp-server | Partial | 5 of 6 | First-party MCP server published by product org (npm, @openai/mcp-extensions, 20280 score). Add live handshake at /.well-known/mcp for full credit. |
| MCP well-known discovery mcp-well-known-discovery | Fail | 0 of 2 | No MCP server URL discovered |
| Public API with reachable endpoints public-api | Pass | 7 of 7 | API described in documentation at https://developers.openai.com but no machine-verifiable API surface confirmed. GraphQL endpoint at https://openai.com/graphql is reachable; |
| Public API/docs linked from homepage public-api-docs | Pass | 3 of 3 | Documentation site found at https://developers.openai.com |
| Developer portal developer-portal | Pass | 6 of 6 | Developer documentation portal found at https://developers.openai.com |
| OpenAPI spec published openapi-spec | Partial | 5 of 7 | OpenAPI spec published on GitHub at https://raw.githubusercontent.com/openai/openai-openapi/main/openapi.yaml |
| API catalog (RFC 9727) api-catalog-rfc9727 | Fail | 0 of 2 | No /.well-known/api-catalog found (RFC 9727) |
| A2A / agent-card a2a-agent-card | Fail | 0 of 2 | No A2A agent card found at /.well-known/agent-card.json |
| Agent discovery file agent-discovery-file | Fail | 0 of 2 | No agent discovery file found |
| OAuth 2.0 support oauth-support | Pass | 5 of 5 | OpenID Connect discovery endpoint found at https://auth.openai.com |
| OAuth Protected Resource metadata (RFC 9728) oauth-protected-resource | Fail | 0 of 2 | No /.well-known/oauth-protected-resource found (RFC 9728) |
| auth.md exists auth-md-exists | Fail | 0 of 2 | No /auth.md found at site root |
| WebMCP support webmcp | Fail | 0 of 5 | No WebMCP support detected - no document.modelContext / navigator.modelContext usage found |
| NLWeb /ask endpoint nlweb-ask | Pass | 1 of 1 | NLWeb /ask endpoint exists but requires authentication (HTTP 403, via GET) |
| CLI tool available cli-tool | Pass | 3 of 3 | CLI package found on npm: @openai/codex |
| Multi-language SDK packages rest-sdk-packages | Pass | 3 of 3 | SDK packages found across 4 ecosystems: npm, pypi, go, rubygems |
| Rate limit response headers rate-limit-headers | Fail | 0 of 2 | GraphQL introspection auth-gated and no header convention documented on GraphQL-tagged pages |
| Idempotency-Key support idempotency-key-support | Fail | 0 of 3 | GraphQL introspection auth-gated and no idempotency documentation found on GraphQL-tagged pages |
| JSON error responses json-error-responses | Pass | 4 of 4 | no detail published |
At the door
What the site answers when a request arrives with an agent's user agent, and what robots.txt says to agents. 2 checks in this group do not apply to this site and are left out.
| Check | Result | Points | What Ora saw |
|---|---|---|---|
| Not blocked by bot detection bot-detection | Partial | 0 of 2 | Bot management challenged ChatGPT-User, OAI-SearchBot, Claude-User, PerplexityBot. A challenge is an identity check that verified crawlers pass, |
| Agent crawler reachability agent-crawler-reachability | Partial | 0 of 2 | Bot management challenged ChatGPT-User, OAI-SearchBot, Claude-User, PerplexityBot - a challenge is an identity check verified crawlers pass, so reachability is undetermined rather than blocked (ChatGPT-User: challenged, |
| robots.txt AI crawler policy robots-ai-policy-quality | Pass | 2 of 2 | AI search crawlers: allowed in 3 of 3 checks (3 crawlers on 1 page). AI training crawlers: allowed in 2 of 2 checks (2 crawlers on 1 page), not scored. |
| robots.txt agent-user policy robots-agent-user-policy | Pass | 2 of 2 | No applicable crawl or usage restriction found for 3 user-triggered agents (chatgpt-user, claude-user, perplexity-user) on 1 public page: https://openai.com/. Declared policy only; |
| Redirect hygiene redirect-hygiene | Pass | 1 of 1 | no detail published |
| Web Bot Auth directory web-bot-auth-directory | Fail | 0 of 2 | No /.well-known/http-message-signatures-directory found |
What an agent can read
Whether the text is in the first response, what the site publishes for agents, and how it describes itself to them. 3 checks in this group do not apply to this site and are left out.
| Check | Result | Points | What Ora saw |
|---|---|---|---|
| Content without JavaScript content-no-js | Partial | 0 of 3 | Homepage blocked by WAF/security challenge - content not accessible to agents |
| Sitemap exists sitemap | Pass | 2 of 2 | Valid sitemap found at https://openai.com/sitemap.xml with multiple sitemaps entries |
| Sitemap freshness (lastmod) sitemap-lastmod | Pass | 1 of 1 | no detail published |
| llms.txt exists llms-txt-exists | Pass | 1 of 1 | Found the llms.txt at https://developers.openai.com/llms.txt. |
| llms.txt formatting llms-txt-formatting | Pass | 2 of 2 | The llms.txt is well-formatted: 47 lines with markdown links, 6,201 characters in total. |
| Markdown agent docs markdown-negotiation | Fail | 0 of 1 | Missing markdown content negotiation. Homepage returned text/html when probed with Accept: text/markdown (expected text/markdown), and none of the well-known root paths (/llms.md, /auth.md, /agents.md, /developers.md, |
| Markdown URL fallback markdown-url-fallback | Fail | 0 of 2 | Homepage does not support the .md suffix. Probed https://openai.com/index.md and it did not return markdown (expected text/markdown content-type or a heading-led non-HTML body). |
| Agent mode view agent-mode-view | Fail | 0 of 2 | No agent mode view detected at ?mode=agent (optional) |
| JSON-LD structured data json-ld | Partial | 1 of 4 | JSON-LD found (1 block(s)) but no identity type - use SoftwareApplication, Product, Organization, Person, or Article so agents can tell what this site is |
| Organization schema completeness org-schema-completeness | Fail | 0 of 2 | No Organization type found in JSON-LD |
| JSON-LD entity linking (sameAs) json-ld-entity-linking | Fail | 0 of 2 | No sameAs entity linking in JSON-LD - agents cannot disambiguate your brand |
| Schema type breadth schema-type-breadth | Fail | 0 of 2 | No extended schema types found - AI can only answer basic questions about this entity |
| Metadata completeness metadata-completeness | Pass | 2 of 2 | All metadata signals present: canonical URL, lang="en", og:image, og:type |
| Agent instruction / when-to-use agent-instruction | Partial | 2 of 3 | When-to-use guidance found at https://developers.openai.com |
Found and recommended
Whether an assistant can find the company by name, confirm what it is, and see it in the registries agents read. 2 checks in this group do not apply to this site and are left out.
| Check | Result | Points | What Ora saw |
|---|---|---|---|
| Brand name discoverability brand-search-accuracy | Partial | 2 of 3 | openai.com ranks #7 for "openai" and appears 3 times in the top 10 (searched your brand name alone: we could not read a product term from your site) |
| Wikipedia / Wikidata entity presence wikipedia-presence | Pass | 4 of 4 | Wikipedia "OpenAI" and Wikidata Q21708200 both verified - domain confirmed on both sources |
| Developer resource discoverability agentic-search-specific | Pass | 3 of 3 | no detail published |
| NPM/PyPI SDK package npm-sdk-package | Pass | 1 of 1 | no detail published |
| ChatGPT app listed chatgpt-app-listed | Pass | 2 of 2 | no detail published |
| Listed on skills.sh skills-sh-listed | Pass | 1 of 1 | no detail published |
| ARD discovery ard-catalog | Fail | 0 of 1 | No /.well-known/ard.json, no same-site rel="ard" link, and no legacy /.well-known/ai-catalog.json |
Payments
The agent payment protocols Ora probes for. A site that sells nothing has nothing here, and Ora scores the layer out of zero when none applies. 5 checks in this group do not apply to this site and are left out.
| Check | Result | Points | What Ora saw |
|---|---|---|---|
| ACP - Agentic Commerce Protocol acp-support | Partial | 0 of 3 | ACP documented (AI analysis: Docs list Commerce flows in ChatGPT; checkout implied); no live protocol surface found |
The fixes worth most points
Ora attaches an estimate to each failing or partial check: how many points the site would gain by passing it. These are the 5 with the largest estimates for OpenAI, with what Ora observed. The wording of each fix is Ora's; the 28 checks it runs on an MCP server are listed with their layer and tier.
- WebMCP support: fail, 0 of 5, about 3.6 points to gain. Ora saw: No WebMCP support detected - no document.modelContext / navigator.modelContext usage found
- MCP Apps support: fail, 0 of 4, about 2.9 points to gain. Ora saw: No MCP Apps support detected - this product has an in-agent UI surface but exposes no ui:// resources or tool UI metadata for agents to render.
- GraphQL typed error model: fail, 0 of 3, about 2.2 points to gain. Ora saw: GraphQL introspection auth-gated and no typed graphql error model documentation found on GraphQL-tagged pages
- Idempotency-Key support: fail, 0 of 3, about 2.2 points to gain. Ora saw: GraphQL introspection auth-gated and no idempotency documentation found on GraphQL-tagged pages
- Scoped permissions: partial, 2 of 5, about 2.2 points to gain. Ora saw: Scoped permissions described at https://developers.openai.com, but no machine-readable scope declaration found (OpenAPI oauth2 scopes or RFC 9728 scopes_supported)
Our read
OpenAI's corporate domain challenges the agents that use it: bot management stopped ChatGPT-User, OAI-SearchBot, Claude-User and PerplexityBot, the home page came back as a security challenge, and the content row fails with it. The developer surfaces live on developers.openai.com and read well from there: an llms.txt, OpenID Connect discovery at auth.openai.com, @openai/codex on npm, SDKs in four registries and an OpenAPI specification that Ora credits from the GitHub repository rather than the domain. The MCP credit is a first-party package on npm, @openai/mcp-extensions, with no live endpoint at a URL Ora could discover, and scopes are described rather than declared. The JSON-LD on the home page has no identity type and the brand search places openai.com seventh for its own name. The company whose crawler is named in every robots.txt in this series is itself behind a challenge page; the checklist's door rows come first here.
Test your own surfaces
The scan above is free and needs no account: enter a domain at ora.ai and read the same four layers for your own site. For an MCP server, Ora scans the server URL itself and runs 28 of its 132 public checks on it, which is why it is the check this directory recommends; the reasons, the limits and the alternatives are on their own page. The command below runs the same audit from a terminal and exits with 1 under the score you set, so it can gate a deploy.
npx ax@0.10 audit https://example.com --min-score 70All 360 checks, one company a day. Corrections to the corrections page.