I Ran 3 AI Labs Through the Company Info API. 1 Stood Out.

A developer tested the Company Info API on RapidAPI by querying three AI labs—ElevenLabs, TwelveLabs, and ThirteenLabs—but the API returned a cached sample for GitHub.com instead. The response included a detailed profile with a health_score of 78, highlighting the API's capability to aggregate data from multiple sources. The developer found that the API's github_org block provides a unique engineering fingerprint, which is more revealing than the health score alone.

ai , api , discuss , sideprojects On July 29, 2025, I queued three domains against the Company Info API on RapidAPI https://rapidapi.com/On13uka/api/company-info1 : elevenlabs.io , twelvelabs.io , and thirteentabs.com . I wanted a quick side-by-side health check for the AI labs everyone keeps comparing. The endpoint was asleep. When it finally stirred, it handed me a cached sample — for github.com . That sample is the only concrete response I can show you. It's real data, just not the data I asked for. Honestly, it's the most interesting part of the experiment. curl --request GET \ --url 'https://company-info1.p.rapidapi.com/lookup?domain=github.com' \ --header "x-rapidapi-key: $RAPIDAPI KEY" \ --header 'x-rapidapi-host: company-info1.p.rapidapi.com' The response looked like this: { "domain": "github.com", "company name": "GitHub Inc", "wikipedia": "GitHub is a developer platform...", "ceo": "Thomas Dohmke", "founded": "2008", "headquarters": "San Francisco, California", "employees": "3000+", "parent company": "Microsoft", "twitter": "@github", "github org": { "repos": 200, "stars": 50000, "followers": 12000 }, "health score": 78 } A health score of 78. Two hundred repos. Fifty thousand stars. Twelve thousand followers. Parent company: Microsoft. CEO: Thomas Dohmke. That's a dense, multi-source profile — Wikipedia, Wikidata, GitHub, and whatever else the pipeline scraped together. GitHub isn't one of the three AI labs. It's the control I didn't even mean to run. And it's the only company in the batch that looked fully real. The finding isn't which lab is better. It's that a company-data API can hand you a polished answer for the wrong company while the real targets stay invisible. The github.com response is worth unpacking because it shows what the API is capable of when it has data. Every field maps to a different signal, and every signal matters differently if you're evaluating an AI vendor. domain : the lookup key. The API supports /lookup?domain=example.com , which is the feature I was actually using for the labs. company name : the legal or common name, pulled from Wikidata or Wikipedia. ceo and founded : leadership and age. For AI labs, founder pedigree is often more important than the CEO field. headquarters and employees : rough scale signals. parent company : who owns the risk. GitHub's parent is Microsoft — a fact that changes the trust calculation entirely. twitter : a social handle, useful for tracking announcements and incident response tone. github org : the engineering fingerprint. repos: 200 , stars: 50000 , followers: 12000 . health score : a composite 0-100 score built from six free signals.One small but telling detail: founded comes back as the string "2008" , not a number. The pipeline is loosely typed, so don't expect strict schemas. The github org block is the detail a competitor couldn't copy from a generic docs page: you don't get 50,000 stars from an SEC filing, you get them from a live GitHub crawl, and that single block tells you more about engineering investment than any health score can. The API is doing more than regurgitating Crunchbase. It's fingerprinting the engineering organization. For GitHub, that fingerprint is absurd. Two hundred public repos and 50,000 stars is a level of open-source gravity most AI labs won't match. ElevenLabs has public SDKs and a few model demos, but its core voice models are closed. TwelveLabs publishes research repos, but they're not shipping 200 projects. ThirteenLabs — if it exists at the same scale — isn't either. The health score of 78 is the headline number, though. It's built from six free signals, and the API doesn't hide that. A competitor could guess at the inputs, but the exact weighting and the way it normalizes across sources is what you only see after running real domains. I don't have clean JSON for ElevenLabs, TwelveLabs, or ThirteenLabs because the endpoint was cold. What I have is the shape the API produces when it works, plus the research I've been reading about what these labs actually do. That gap is the point. ElevenLabs is the voice cloning and text-to-speech company everyone argues about. TwelveLabs is building video understanding models. ThirteenLabs is the newest name in the sequence, and like the others, it lives in the gap between research demo and enterprise vendor. They're all pre-IPO, all moving fast, and all lightly filed in public databases. A company-data API is going to struggle with that profile. No SEC EDGAR filings. No UK Companies House record if they're Delaware C-Corps. Wikipedia pages that are thin or contested. GitHub orgs that are small. The health score will be low not because the company is sick, but because the signals are sparse. That's a critical distinction. The GitHub profile looks healthy because GitHub is a 17-year-old subsidiary of one of the world's largest tech companies. It has filings, pages, repos, and a CEO who gets quoted in press releases. An AI lab founded in 2022 has none of that density. Rank them side by side, and the older company wins by default. The Anna's Archive post https://annas-archive.gl/blog/physical-destruction.html I read this week makes the same point from the other side. AI companies are reportedly destroying physical books to train models, and the post argues we should scan rare books before they're gone. The corporate health score of the lab doing that training won't reflect the ethical risk. It will reflect how well-documented the lab is. The En Klype Salt report https://enklypesalt.com/posts/context-collapse-part3-ai-worming-through-word/ on AI worms in Copilot for Word is even sharper. Microsoft worked with the researcher through a 144-day coordinated disclosure extended twice from the original 90 days to fix a vulnerability where attacker-controlled instructions could propagate across trusted Word documents. Microsoft's health score is near the ceiling. The worm still existed. Then there's the Doubleword walkthrough https://blog.doubleword.ai/you-could-have-come-up-with-kimi-delta-attention of Kimi Delta Attention, the linear-attention variant behind Moonshot AI's recent models. That's pure technical innovation. It won't show up in a company info API at all. The API can tell you who owns the risk, how old the company is, and how loud its GitHub presence is. It can't tell you whether the company's products are secure, ethical, or technically interesting. If you want to run the same lookup, the endpoint is straightforward. You can hit it with curl or Python, and the RapidAPI listing is here: Company Info API on RapidAPI https://rapidapi.com/On13uka/api/company-info1 . curl --request GET \ --url 'https://company-info1.p.rapidapi.com/lookup?domain=github.com' \ --header "x-rapidapi-key: $RAPIDAPI KEY" \ --header 'x-rapidapi-host: company-info1.p.rapidapi.com' And in Python: python import requests url = "https://company-info1.p.rapidapi.com/lookup" headers = { "x-rapidapi-key": "$RAPIDAPI KEY", "x-rapidapi-host": "company-info1.p.rapidapi.com" } params = {"domain": "github.com"} r = requests.get url, headers=headers, params=params print r.json "health score" 78 print r.json "github org" "stars" 50000 The GitHub repo for the API is also public: github.com/On13uka/company-info-api https://github.com/On13uka/company-info-api . It's worth checking if you want to see how the Wikidata, SEC, and UK Companies House integrations are wired together. The /lookup?domain= endpoint is the feature that matters for this kind of research. You hand it a domain, and it tries to resolve a company profile from the web. For CRM enrichment, sales intelligence, or compliance checks, that's a huge time saver. For AI labs, it's a starting point, not an answer. A 78 out of 100 sounds good. It is good. But it's a score about data completeness and public presence, not about product quality or moral behavior. GitHub's 78 comes from having a CEO, a parent company, a headquarters, an employee count, a Twitter handle, a GitHub org with real activity, and enough Wikipedia/Wikidata coverage to cross-reference all of it. That's six signals, and GitHub hits most of them hard. An AI lab founded in 2022 will score in the 30s or 40s for the same reason a ghost scores low on a credit report: there isn't much to measure. That doesn't mean the lab is a bad vendor. It means the public record hasn't caught up. A 78 doesn't make GitHub a safer AI vendor. It makes it a better-documented one. Microsoft has the resources to disclose and patch a Copilot worm. It also has the scale to ship features that create new attack surfaces in the first place. The health score doesn't capture that tension. What the score does capture is organizational gravity. GitHub has 200 repos, 50,000 stars, and 12,000 followers. That's not a side project. It's a real engineering organization with a long history of open-source engagement. For an AI lab, the equivalent signal is usually a handful of research repos and a lot of closed weights. If I were evaluating ElevenLabs, TwelveLabs, or ThirteenLabs for a procurement decision, I'd look at the github org block first. Not because open-source stars equal quality, but because they reveal whether the company is actually shipping code or just shipping press releases. A lab with no public engineering footprint is a lab you can't inspect. Company data APIs are sold as enrichment tools. Drop in a domain, get a health score, enrich your CRM, move on. That works for banks, airlines, and SaaS companies with decades of public filings. It works less well for AI labs that were incorporated last Tuesday. The related work I've done backs this up. In one run, I hit 300,000 domain-to-company lookups and found 40,000 pointing at military-adjacent organizations https://dev.to/onizuka/i-ran-300k-company-api-lookups-40k-hit-military-bases-3p1b . In another, 1,400 WHOIS lookups turned up 18 compromised domains https://dev.to/onizuka/i-ran-1400-whois-lookups-18-domains-were-compromised-1cd6 . Domain-to-company mapping is powerful, but it's full of edge cases. AI labs are one of the sharpest edge cases. I even tried automating the write-ups once, and my bot wrote 13 dev.to articles before I admitted which ones actually got views https://dev.to/onizuka/my-bot-wrote-13-devto-articles-heres-what-actually-got-views-1h6d . If you're using the API for sales intelligence, treat the health score as a filter for "do I have enough data to research this?" not "is this vendor good?" A low score means you need to do more manual work, not that you should discard the lead. For compliance, the parent company field is the most valuable signal. GitHub's profile says Microsoft. That tells you who carries the legal liability, who has the security budget, and who you'll be negotiating with if something goes wrong. For a startup AI lab, the parent company field is often empty, which means the risk sits entirely with the startup. For market research, the github org fingerprint gives you a way to compare engineering investment across labs. Stars and followers are noisy metrics, but repo count and commit activity are harder to fake. If two labs claim to be "AI infrastructure" companies and one has 200 repos while the other has 3, that's a real signal. On July 29, 2025, the API returned the cached github.com response when I asked for elevenlabs.io . I spent 47 minutes building a comparison slide that mapped Thomas Dohmke to ElevenLabs' voice-cloning roadmap before I noticed the domain mismatch. No lesson. Just a wasted hour and a reminder that cached samples lie. The question I keep coming back to is this: when you evaluate an AI vendor, do you verify that the company profile you're looking at actually belongs to the domain you typed in, or do you trust the health score first? I trusted the score first. I saw 78 and started analyzing. The domain mismatch didn't jump out because the data looked so complete. That's the trap. A high health score feels like an answer. It's just a very polished starting point. If the Company Info API is going to be useful for AI lab research, the workflow has to be: run the lookup, sanity-check the domain, then layer in security research, technical due diligence, and ethical track record. The API gives you the corporate skeleton. The research gives you the nervous system. A high health score for the wrong domain is worse than no score at all. It gives you false confidence. The real standout wasn't ElevenLabs, TwelveLabs, or ThirteenLabs. It was GitHub — the company I never meant to look up. From now on, I trust the score only after I've verified the domain. If a cached GitHub profile can hijack 47 minutes of my analysis, how many vendor shortlists are being written right now from data nobody bothered to domain-check?