Dr7.ai
Dr7.ai is a medical AI hub routing one account and one API key to 15+ clinical models, through either a browser diagnosis interface or a developer API. Its own terms restrict use to research and education, not clinical diagnosis.
What is Dr7.ai?
Dr7.ai bills itself as "The World's Advanced Medical AI Hub" and "the unified API gateway for all medical AI models worldwide". The premise is aggregation: rather than onboarding separately with every publisher of a clinical model, you open one account and reach a catalogue the homepage advertises as 15+ models (its own pages also say "14+ Models" and "13+ specialized", so the count is not stable). The models named on the site include MedGemma, Med-PaLM 2, BioGPT, Meditron, PMC-LLaMA, CheXagent, LLaVA-Med, Med-Flamingo, BiomedLM, BiomedCLIP, Clinical Camel, MedSigLIP, AlphaGenome and Baichuan-M3, sorted on the /models page into Genomics, Text Analysis, Vision and Multimodal, and credited by the site to Google, Google DeepMind, Stanford CRFM and Baichuan.
There are two doors. The first is a no-code consultation interface: you pick a model family (All, Text, Expert, Vision or Multimodal), type symptoms, and optionally upload a medical image. The second is the API. The documentation is concrete: a POST to /v1/medical/chat/completions, an Authorization: Bearer sk-dr7-... header, a chat-completion response shape with choices, usage and prompt_tokens, plus endpoints for listing models, streaming completions and image analysis, and documented error codes (401 invalid key, 402 insufficient balance, 429 rate limit, 500 server error). Pricing is per token, published model by model, from 0.001 to 0.006 USD per 1K. The selling line is "Integrate once, switch between models anytime without multiple onboarding processes". The homepage advertises EHR analysis, imaging interpretation, drug discovery and medical Q&A, and the site is deployed across 13 locales, though the French pages show only partial translation.
One frame governs all of it. The terms, section 5, prohibit clinical diagnosis, treatment planning, emergency situations, patient care decisions, pharmaceutical recommendations and processing of real patient data, and a Medical Disclaimer repeated in every footer states the models are "not clinical-grade tools". The headline nonetheless reads "AI-Powered Diagnosis for Doctors". Readers should treat the terms, not the headline, as the binding text.
What it does
- Query several medical language models from one interface, without a separate account per publisher
- Upload and analyse a medical image (X-ray, CT, MRI) through a vision or multimodal model
- Call the same models from your own code via a single API key
- Compare which models each plan unlocks, and what each costs per 1K tokens in and out
- Keep, tag and retrieve past consultations (paid plans only)
When to use Dr7.ai / When not to
A quick filter to help you decide if Dr7.ai is the right fit.
When to use Dr7.ai
- Developers who want to try several medical language and vision models behind a single API key instead of onboarding with each publisher separately.
- Medical students and thesis writers, the audience the site itself names for its Standard plan.
- Educators and teaching teams, the audiences named for the Pro and Clinical+ plans, building teaching material or case libraries.
- Biomedical researchers exploring the literature (BioGPT) or genomic questions (AlphaGenome) within a research-only scope.
- Technical teams prototyping or benchmarking medical models before committing to one vendor, using the published per-token rates to estimate cost.
When not to use Dr7.ai
- Clinicians looking for a real diagnostic, triage or treatment-planning tool: Terms of Service section 5 explicitly prohibits clinical diagnosis, treatment planning and patient care decisions.
- Anyone facing an emergency medical situation, which the same section 5 lists among the prohibited uses.
- Teams that need to process identifiable patient data: section 3 makes the user responsible for anonymising everything submitted, and the privacy policy says no patient database is kept.
- Organisations that require a legally identified vendor, a signed DPA or evidence behind a certification claim, none of which the site provides.
- Under-18s, and commercial medical operators without explicit authorisation, both excluded by the terms.
How to use Dr7.ai
A typical end-to-end flow, from setup to results.
- Decide which door you need: the no-code diagnosis interface, or the developer API.
- Create an account. Sign-up runs through Google, no credit card is required, and the homepage advertises about 2 minutes.
- Pick a plan, from the permanent free tier to the paid tiers that bundle monthly API credits; the homepage advertises the step as "Instant".
- No-code path: open the diagnosis page and choose a model family (All, Text, Expert, Vision or Multimodal).
- Type symptoms or upload a medical image. On the free plan you get 3 demo text messages and 1 image analysis per day, with no API access.
- Developer path: generate an API key from the dashboard; keys carry the sk-dr7- prefix.
- Send a POST to https://dr7.ai/api/v1/medical/chat/completions with model, messages, max_tokens and temperature; the docs use "medgemma-4b-it" as the sample model.
- Authenticate with the Authorization header, in the form "Bearer sk-dr7-" followed by your key, and follow the documentation's instruction never to expose that key client-side.
- Handle the documented error codes: 401 invalid key, 402 insufficient balance, 429 rate limit, 500 server error.
- Anonymise anything you submit, which section 3 of the terms makes your responsibility, and keep to the research and education scope set by section 5.
Pros & Cons
Pros
- One access point and one key for 15+ medical models published by different organisations, which is the whole point of the product
- API pricing published openly, per token and model by model, in and out per 1K
- A clear grid showing which models each plan unlocks, published before you pay
- A permanent free plan that needs no credit card, enough to see the interface work
- Concrete API documentation: endpoint, authentication header, curl sample, response sample and error codes
- A diagnosis interface usable without writing any code
- 13 locales deployed, and a medical disclaimer repeated in the footer of every page
Cons
- No identifiable operator: no company name, no address, no legal notice and no contact page anywhere in the site's 286-URL sitemap, only "(c) 2025 Dr7.ai" in the footer
- The Privacy Policy and Terms of Service are written in the name of "MedGemma", not Dr7.ai, and both carry "Last updated: December 2024" - seven months before the dr7.ai domain was registered on 20/07/2025
- The only usable contact addresses published are on a different domain: sales@medgemma.org on the pricing page and support@medgemma.org in the terms; sales@dr7.ai appears only on the API access page
- SOC 2 Type II, ISO 27001, HIPAA, GDPR and PIPEDA are claimed with no supporting evidence: no trust page, no named auditor, no report
- Headline figures are given without a source - "Active in 50+ hospitals", "95% accuracy rate", "Join 1000+ medical professionals", "99.9% Uptime", "<100ms Response" - and the diagnosis-for-doctors pitch contradicts the terms, which forbid clinical diagnosis and care decisions
- No DPA, no subprocessor list, no EU Article 27 representative, no GDPR contact address, no documented way to exclude your data from training, and no hosting country beyond the name "Google Cloud"
- Refunds are limited to 2 calendar days after the subscription starts, and there is no social profile, team page or blog reachable from the sitemap
Pricing & Plans
A permanent free plan is available and requires no credit card. The lowest paid entry point is 9.99 USD per month (Standard), followed by Pro at 49.99 USD per month, currently displayed at 34.99 USD per month under a 30% promotion, and Clinical+ from 199.99 USD per month. Annual billing is advertised at "Save 20%". Paid plans include monthly API credits of 5, 25 and 100 USD respectively; beyond those credits, API usage is billed per token, from 0.001 to 0.005 USD per 1K input and 0.002 to 0.006 USD per 1K output depending on the model. A full refund may be requested by email within 2 calendar days of the subscription start, processed in 5 to 7 business days. Enterprise sales are handled by email only.
- 3 text diagnosis messages
- 1 image per day
- no API access
- no multi-turn dialogue
- no history.
- 5 USD of API credits
- MedGemma-4B
- unlimited text
- up to 20 images per day
- 5 dialogue turns per day
- recent history.
- 25 USD of API credits
- all models (MedGemma-4B and 27B)
- unlimited text and images
- unlimited dialogue
- extended history with collections and tags.
- 100 USD of API credits
- all models plus priority access and future models
- batch upload
- conversation logs and collaboration
- shared searchable case library.
Data, GDPR & hosting
A consolidated view of how Dr7.ai handles your data.
GDPR overview
The site claims GDPR compliance in so many words: a "HIPAA / GDPR ready" badge on the homepage, and a privacy policy stating it "adheres to international healthcare data protection standards including HIPAA (US), GDPR (EU), PIPEDA (Canada)". It lists rights: immediate deletion, access to processing logs, portability, restriction of processing. That claim is the entire basis for the "Y" recorded here, and it is a self-declaration, not a verification. Nothing supports it: no legal basis for processing is stated, no DPO is named, no GDPR contact address and no Article 27 EU representative exist, no DPA is published or offered, no subprocessor list is available, and no hosting country or region is specified. Requests must go through "secure contact channels" the page never names. The policy is written in the name of "MedGemma", not Dr7.ai.
Who owns the data?
The terms never say who owns what. Nothing grants the user ownership of submitted content, and nothing claims it for the operator. Section 3 places the anonymisation duty on the user: "Users are responsible for ensuring submitted medical data is anonymized and complies with applicable privacy laws (HIPAA, GDPR, etc.)". Section 4 reserves the interface, documentation and educational content to the publisher and forbids reverse-engineering or redistributing the models, while allowing AI responses to be used for non-commercial research and education with attribution. The privacy policy adds that queries are processed anonymously and states: "We do not store personal health records, maintain patient databases, or provide clinical decision support." No legal entity is ever named as the counterparty to any of this.
Reuse rights
On the user's side, section 4 of the terms allows AI responses to be reused for non-commercial research and education with attribution; commercial medical use requires explicit authorisation, and the models themselves may not be reverse-engineered or redistributed. On the operator's side, the privacy policy declares it collects anonymised medical queries, diagnostic images without identifiers, model interaction logs and technical metrics, and states that data is "used exclusively for AI model inference, performance analysis, and platform improvement - never for commercial purposes or third-party sharing". What "platform improvement" covers is never defined: the policy never says, either way, whether customer data feeds model training, and no opt-out is documented anywhere on the site. No subprocessor list is published; Google Cloud is the only infrastructure named.
Data retention & training
Hosting summary
The only infrastructure named is Google Cloud: the privacy policy states that "Your medical data is processed using Google Cloud's secure infrastructure with advanced encryption, access logging, and compliance monitoring." No hosting country and no region are specified anywhere, so the jurisdiction that would apply to the data is undetermined. The declared security measures are end-to-end encryption, secured API endpoints, audit trails, access logging and role-based access controls, under a framework the site describes as SOC 2 Type II, ISO 27001 and Google Cloud's healthcare compliance programme - all claimed, none evidenced by a report, an auditor or a trust page. On the network side, the domain resolves to 188.114.96.12 on AS13335 (Cloudflare, anycast), which identifies the CDN in front of the site and says nothing about where processing actually happens. No subprocessor list is published and no DPA is offered.
Things to keep in mind
Risks and trade-offs to weigh before adopting Dr7.ai.
- No identifiable publisher. The 286-URL sitemap contains no legal notice, no about page and no contact page, and no company name or address appears anywhere. If something goes wrong - a billing dispute, a data request, a liability question - there is no named party to address it to.
- The legal pages belong to someone else's site. Both the Privacy Policy and the Terms of Service are written in the name of "MedGemma", the name of an open model published by Google DeepMind, and both carry "Last updated: December 2024", seven months before the dr7.ai domain was registered on 20/07/2025. The two published contact addresses, sales@medgemma.org and support@medgemma.org, are on that other domain too.
- Heavy certifications are claimed with nothing behind them. SOC 2 Type II, ISO 27001, HIPAA, GDPR and PIPEDA appear in the privacy policy without a trust page, a named auditor or a report. A reader who buys on the strength of those words is buying a sentence, not an audit.
- Adoption and performance figures carry no source: "50+ hospitals", "95% accuracy rate", "1000+ medical professionals", "99.9% Uptime", "<100ms Response". A "30% OFF - Limited Time Offer" banner sits on every page collected, with no end date, so the urgency it signals is not real urgency.
- The pitch and the contract point in opposite directions. The homepage sells diagnosis to doctors and claims hospital deployment, while section 5 of the terms forbids clinical diagnosis, treatment planning, emergency use and care decisions. The risk is human, not technical: a plausible, fluent, confidently worded answer invites a tired clinician or an anxious student to lean on it, and automation bias does the rest. Any output here needs independent verification before it touches a decision.
- You carry the data duty alone. Section 3 makes anonymisation your responsibility, no DPA is offered, no subprocessor list is published, no hosting country is named beyond "Google Cloud", and no opt-out from training is documented while the policy reserves the right to use data for undefined "platform improvement".
- The site resells access to models, including Med-PaLM 2, without documenting a licence or a redistribution agreement, and the actual availability of every billed model has not been demonstrated on the collected pages.
Setup & Integrations
Technical difficulty
Two very different efforts. The web route requires no technical skill at all: sign in with Google, open the diagnosis page, pick a model, type symptoms or upload an image. The site advertises about 2 minutes for the account and "Instant" for the plan. The API route is standard developer work: one HTTP POST with a Bearer header and a chat-completion payload, which anyone who has called an LLM API will recognise. No official SDK or client library is offered, and no webhooks or third-party integrations are announced, so any orchestration is yours to write.
Deployment
Supported languages
Behind Dr7.ai
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
Can I use Dr7.ai to diagnose a patient?
How much does Dr7.ai cost?
Is there an API?
Which models can I reach through it?
Who publishes Dr7.ai?
How long is my data kept?
Is Dr7.ai GDPR compliant?
Can I get a refund?
Is there a minimum age?
Is there a mobile app?
Should you pick Dr7.ai?
On the evidence, Dr7.ai is a deployed product rather than a landing page: 286 URLs, 13 locales, a four-tier price list with per-token rates published model by model, API documentation with an endpoint, an authentication header, a curl sample, a response sample and error codes, and a diagnosis interface you can open. As a bench for trying several medical language and vision models behind one key, it is legible and cheap to test - the free tier asks for no card.
What it does not provide is anyone to hold to it. No company name, no address, no legal notice and no contact page appear in the whole 286-URL sitemap; the footer carries only "(c) 2025 Dr7.ai". The Privacy Policy and Terms of Service are written in the name of MedGemma, not Dr7.ai, and both are dated "Last updated: December 2024" - seven months before the domain was registered on 20/07/2025. The only usable contact addresses sit on another domain, medgemma.org. SOC 2 Type II, ISO 27001, HIPAA, GDPR and PIPEDA are claimed without a trust page, an auditor or a report, and "50+ hospitals", "95% accuracy rate" and "1000+ medical professionals" are stated without a source.
The gap that matters most is internal to the site. The headline sells "AI-Powered Diagnosis for Doctors" while section 5 of the terms prohibits clinical diagnosis, treatment planning, emergency use and patient care decisions. The contract, not the headline, is what binds a user. Read that way, Dr7.ai is usable for research, teaching and model comparison, and out of bounds for anything touching a real patient. For any commitment beyond a personal test - a purchase order, a DPA, real data - the missing publisher identity is the first thing to establish.
- Choosing a selection results in a full page refresh.
- Opens in a new window.