Open Wearables - Open infrastructure for wearable-powered health products.

@kaliszs Fair take Sebastian, and honest. MCP was hyped as the universal answer and it's clearly not, the spec is still moving, the tooling is uneven, and "agent reads your data" is harder in practice than the demos suggest. That said, for a narrow use case like structured health context for an LLM, it's still the cleanest pattern we've found. Worth shipping, not worth overselling.

@kaliszs Yeah, that tracks with how a lot of infrastructure-first ideas evolve in practice.

@kaliszs haha yeah the origin story is a bit ironic

MCP had a lot of hype and reality is more nuanced. the protocol is useful but it's not magic - the hard part is still the context you feed it, which is exactly why the data and scoring layers matter

the server will get there. the foundation is what makes it worth building on

@gregory80pl Garmin is fully supported. Deploy on Railway in about 5 minutes and your watch data starts flowing through the API. GitHub link is in the description if you want to dig in.

@gregory80pl We too xd Garmin is a tough one, but we handled it!

@ajaykumar1018 You can run the repo locally and setup the MCP server. Let us know your thoughts!

@ajaykumar1018 Feel free to use our MCP, but remember - it's still not 100% production ready.

@eva_kuttichova2 Yes we have both Flutter and React Native SDK.

@eva_kuttichova2 Btw, you find source code of both on our Github. Invitation links to try them are located in one of the Discord's channels ;)

@jakub_mikolajczyk1 Thanks Jakub, that's exactly the use case we built it for. Six months of OAuth flows, schema mapping, and provider edge cases is real time most teams don't have. If your friend wants to compare notes (or just complain about Garmin's docs), send her our way: Discord is the fastest path, or DM works too.

@jakub_mikolajczyk1 High five! And thanks!

@jakub_mikolajczyk1 six months is honestly conservative for some teams - the provider integrations alone are brutal to maintain

tell her we're here if she wants to talk through the use case: https://discord.gg/openwearables

@joanna_samek Thanks Joanna! Honest answer: closer to "low maintenance" than "set and forget."

Once it's running, the main things are pulling updates when we ship a release, watching logs for errors, and occasional database housekeeping. Provider APIs change sometimes (Garmin tweaks something, Oura updates a field), we patch it, you pull the update. Security patches are mostly automated via Dependabot.

Realistically, a small team spends a few hours a month on it. If that sounds like more than you want to take on, Momentum offers managed deployments where we handle all of that for you.

@joanna_samek the easiest option is to use one-click deployment on Railway so you don't have to worry about infrastructure so much. At least with relatively low scale this might be the best solution but once your user base grows it might become a bit pricey

@priyankamandal Thanks Priyanka. Self-hosted + MIT removes two of the hardest conversations with compliance teams: where does the data live, and what happens if the vendor disappears. The architecture is HIPAA-ready out of the box (encryption, audit logging, RBAC), though the cert itself sits with whoever operates the deployment, not the software.

@priyankamandal Open source community is a real strength! We have 40+ contributors.

@michal_kukul sure, probably the easiest way to give it a try is using one-click deployment on Railway - you can check how it works in our docs: https://openwearables.io/docs/deployment/railway

@imtiaj_ahmad Very scalable: FastAPI + PostgreSQL + Redis + Celery scales horizontally, and multiple companies run it in production today. The real advantage at scale is the cost model: zero per-user fees, so where SaaS aggregators charge $0.50-2.00 per user per month, you only pay for your own infrastructure. The bigger you get, the better the math works.

@richieoh Agreed Richie, the Bluetooth SIG profiles (HRP, GHS) had the right idea on paper but vendors have zero incentive to converge, differentiation and lock-in are part of the business model, so a pragmatic abstraction layer is the only thing that ships in our lifetime.