Five launches from August 20: AI agents, health records, and a smarter notebook

If the launch feed yesterday felt like it was all AI agents, you weren’t imagining it. Three of the five products that shipped on August 20 are explicitly built for the agent era. The other two lean on AI in ways that would have been unthinkable a few years ago. The new developer tools landing yesterday cover infrastructure, developer experience, and consumer apps, and the mix is worth a closer look.

One note before diving in: community rankings are still settling, so there’s no definitive top pick today. If I had to bet on which products will climb the charts, Shepherd Terminal and monolog have the broadest immediate appeal.

Hubble

Hubble wants to be the layer that finally unblocks patient data for AI applications. Right now, pulling a complete medical history means dealing with each provider’s proprietary system. Some offer APIs. Many don’t. The industry standard is still phone calls and faxes, which is absurd in 2026.

Hubble’s approach is more aggressive than most. It connects to EHRs and HIEs directly when possible, but it also deploys browser agents to navigate provider portals and voice agents to call records departments. Voice agents making phone calls on behalf of developers is a level of automation most health data startups avoid entirely.

The consent model is worth paying attention to. Users verify their identity once, then can review and revoke access at any time. Given how sensitive this data is, that’s not just a compliance checkbox. It’s the right way to build trust in a system that asks people to hand over their entire medical history.

Who should care: builders creating AI health assistants, researchers needing aggregated data, and anyone tired of acting as their own medical courier. The free trial lowers the barrier to testing it against your own use case.

Shepherd Terminal

If you’ve run more than one AI coding agent at a time, you know the pain. Terminal windows multiply, you lose track of which agent is waiting for input, and accidentally closing the wrong window kills an hour of work.

Shepherd Terminal solves this with persistence. Sessions stay alive even when the app closes. Agents keep running on remote machines even if your SSH connection drops, thanks to a lightweight Rust service with a session keeper. That’s the piece that makes the whole thing work.

The real-time status tracking is a small touch that makes a big difference. Seeing which agents are working, which are blocked, and which are done gives you back the mental bandwidth you were spending on context-switching.

One caveat: it’s macOS only, requires Apple silicon, and your agents need to integrate with it. If you’re already on Codex or Claude, this is worth a try. The free price point makes testing painless.

Finch

Finch sits on the patient side of the same healthcare data problem Hubble tackles. Where Hubble gives developers an API, Finch gives individuals a place to see everything at once.

The app consolidates records and test results, transcribes medical appointments, and translates clinical jargon into plain English. For anyone managing a pregnancy, which the product explicitly calls out, that’s a lifeline. A dozen plus appointments, dozens of results, and a vocabulary that feels like a foreign language.

Appointment transcription is the standout feature. Being able to review what your doctor actually said, without relying on memory, changes how you prepare for follow-up visits.

The honest critique: iOS only limits its reach, and the $99/year Finch+ tier puts the best features behind a paywall. But for expecting