code-blue.ai

What It Is

code:blue is a passive, on-device AI tool for early stroke screening from Code Blue. Its goal is to detect early signs of stroke before patients notice something is wrong, using existing phones, laptops, or webcams in the room, then route help requests to trusted contacts, caregivers, and emergency services. The website emphasizes that it is at the Seed 2026 stage and is seeking partnerships with investors, healthcare systems, insurers, and device manufacturers.

Core Capabilities and Use Cases

The product’s biggest selling point is that it “requires no new hardware”: it uses existing cameras, microphones, and on-device neural engines for quiet monitoring. Its AI models are said to be trained on real stroke imagery obtained in collaboration with UCSF Health, and each detection model is reviewed by clinical advisors before deployment. When a possible stroke is detected, the system notifies the user’s designated family members or caregivers and contacts emergency services, while sharing location and status information. Typical use cases include home monitoring for high-risk individuals, pre-hospital triage, insurance cost control, and integration of health features by device manufacturers.

Pricing, Privacy, and Integration

The website does not disclose pricing, free trials, subscription models, or payment methods, nor does it provide API documentation. On the integration side, it only mentions support for phones, laptops, and webcams, while inviting healthcare systems, insurers, and device manufacturers to run joint pilots. In terms of privacy, code:blue explicitly states “Private by default,” saying that data will not leave the device without permission. When a suspected stroke occurs, users can choose who to notify. This design is well suited to sensitive medical-data scenarios, but the specific compliance framework, data retention policies, and audit mechanisms are not explained.

Pros, Cons, and Limitations

Its strengths are that it targets the critical treatment window for stroke, addresses a clear need, and requires no additional hardware. The team’s background spans Apple/Samsung health features, AI, neurology practice, and FDA regulation. The website also discloses that a small UCSF pilot with n=30 achieved 100% detection accuracy, and that Kaiser Permanente has requested a follow-up study with 1,000 participants. The limitations are also clear: n=30 is still insufficient evidence, and the company has not disclosed false-positive rates, false-negative rates, target populations, FDA approval status, or real-world performance data. If a medical emergency product has not completed regulatory validation, it should not be treated as a mature tool.

Who It’s For and Access from China

At this stage, it is better suited for healthcare institutions, insurers, device manufacturers, and investors evaluating pilots or partnerships, rather than for immediate purchase by general users. There is no public information on access from China, Chinese-language support, or compatibility with local emergency-response systems. Even if the website is accessible, medical-data compliance, emergency-service integration, and payment workflows may require local alternatives or deep localization.