Trio

Trio is an open-source automated insulin delivery (AID) system that runs entirely on an iPhone — there is no Android version.¹ It is built on the OpenAPS algorithm (the "oref" decision-making engine first developed for the original DIY rigs), adapted for iOS.¹ Its lineage runs from Ivan Valkou's FreeAPS X, through the community project later known as iAPS, to Trio, reusing the LoopKit toolkit and a broad set of community-maintained pump and CGM drivers.⁷ When development of iAPS diverged over questions of open-source governance and peer review, the community split off Trio as a separate project, which is described as the recommended landing spot for people migrating off iAPS to an oref-based loop on iPhone.⁸

Algorithm. Like its OpenAPS roots, Trio is described as predicting glucose and micro-dosing insulin every few minutes. It offers super-micro-boluses (SMB), automatic sensitivity adjustment, dynamic insulin sensitivity (Dynamic ISF), and unannounced-meal detection — so it can blunt a rise even when you forget to enter carbs.⁸ In April 2026 the team shipped v0.7.0, described as the first fully stable release of a completely rewritten app, with the whole logic and data layer rebuilt for speed and stability.⁹

Hardware. The current v1.0 release supports Omnipod 5 alongside its existing pump families. New CGMs include Eversense E3/365 and Accu-Chek SmartGuide; direct Libre 3/3 Plus pairing remains beta, reads every five minutes and does not send data to LibreView or LibreLinkUp. Enlite support was removed.⁶

Customizability and community. Because it is fully open source, every setting and the algorithm itself are inspectable and tunable.¹ It has its own documentation site and active community support, and is backed by the Nightscout Foundation organization.¹

Setup and legal status. You build the app yourself — either a "browser build" using GitHub, or a Mac build using Xcode.² A free Apple Developer account requires rebuilding the app every 7 days.² Critically, Trio "has not been evaluated by any regulatory agency and is not approved by any healthcare authority. You build and run Trio at your own risk."¹

Clinical evidence. Trio itself has not been studied in a randomized trial, but it implements the same OpenAPS algorithm tested in the landmark CREATE trial. According to PubMed, that multicenter randomized controlled trial in 97 children and adults found open-source AID raised time-in-range from 61% to 71% over 24 weeks — about 3 hours 21 minutes more per day in range than a sensor-augmented pump — with no severe hypoglycemia or ketoacidosis.¹⁰ A 24-week continuation phase (extending total use to about 48 weeks) confirmed sustained benefit across different pumps and ages with no new safety concerns,¹¹ and a qualitative sub-study reported empowerment and improved wellbeing among Indigenous Māori users.¹² This is the strongest evidence base any open-source loop can point to, though it tested an AndroidAPS/OpenAPS build, not Trio specifically.¹⁰

September release. Version 1.0 replaces the JavaScript dosing implementation with Swift; the maintainers report matching decisions in their replay tests, not new clinical-trial evidence. The release rebuilds alarms and asks users to review notification settings after updating. Read the version-specific migration instructions before changing an existing system.⁶