Loop (iOS)

Loop is an open-source automated insulin delivery (AID) app that runs on the iPhone, built and maintained by the LoopKit community.⁴ Unlike commercial AID, you compile it yourself from source onto your own device, which is the central fact of using it.⁴

Platform and algorithm. Loop is iOS-only; there is no Android version — the equivalent Android project is the separate AndroidAPS.² Its algorithm is a model-predictive control (MPC) design that, every five minutes when new glucose data arrives, generates a new prediction using your therapy settings, modified by active overrides.¹ Both bolus recommendations and temporary basal rate adjustments are set based on this prediction.¹

Supported hardware. Loop 3.14.7 added Omnipod 5 alongside Eros/DASH, Medtrum Nano and legacy Medtronic pumps, with 3.14.8 (20 September 2026) and 3.14.9 (5 October 2026) refining Pod Bluetooth handling, multi-user household Omnipod DASH pairing, Eversense + Omnipod BLE wake-up handling, and Medtrum patch status. Eversense E3/365 support arrived in 3.14.2; Dana remains on a feature branch and Libre 3 is not supported. Read the current compatibility and migration notes before updating.⁶⁷

Community and support. Loop has unusually deep documentation (LoopDocs) and a large volunteer support network offering build help.¹¹⁴ More than 9,000 people were estimated to be using Loop by the time of a 2022 study.¹¹

Legal and self-responsibility status. DIY Loop is not a regulated medical device; you build and run it at your own responsibility, and LoopDocs is explicit that you must know how to set up, operate and troubleshoot it.⁴ Importantly, the Loop algorithm became the basis for Tidepool Loop, which the FDA cleared in January 2023 as the first interoperable automated glycemic controller (iAGC) supported by real-world evidence rather than a classical trial — the first regulatory clearance of an open-source AID algorithm.¹²¹³

Clinical evidence. In a prospective real-world study of 558 adults and children, Loop raised time-in-range from 67% to 73% and lowered HbA1c from 6.8% to 6.5% over six months, with no increase in severe hypoglycemia.¹⁴ Qualitative work with Loop users reports reduced management burden and better quality of life, alongside real setup and troubleshooting friction.¹¹ (The randomized CREATE trial of open-source AID used the OpenAPS algorithm in AndroidAPS, not Loop, so it speaks to the open-source approach generally rather than to Loop specifically.¹²)

More recent work has started to place Loop next to commercial AID and to extend it to children. A 2026 study compared glucose management between open-source Loop and Tandem Control-IQ in adults with T1D — we cite it here only as evidence that such a head-to-head comparison now exists, because it was published as a letter without an abstract, and we will not put numbers on it we have not read.⁹ Separately, a single-centre study followed 50 children aged 2–18 who started Loop: over six months time-in-range rose from 68.3% to 72.7%, the share of children above the 70% time-in-range target rose from 46% to 66%, and HbA1c fell from 7.0% to 6.6% (in a smaller subset with HbA1c data), with no change in time spent low and no episodes of diabetic ketoacidosis.¹⁰ It is one centre and there is no control group, so read it as encouraging rather than definitive.

September–October updates. Loop 3.14.7 (15 September 2026) introduced Omnipod 5 support, followed by maintenance releases 3.14.8 (20 September 2026; Dexcom G7 scan-for-new-sensor toggle and bolus screen state fixes) and 3.14.9 (5 October 2026), which fixed multi-user household Omnipod DASH pairing collisions, resolved an Eversense + Omnipod BLE background wake-up halt, updated MedtrumKit patch status display, and noted that newly shipped "black-dot" Omnipod 5 pods cannot have pod-fault or expiration beeps silenced. These are software maintenance changes, not new clinical outcome evidence. Dana support continues on feature branches.⁶⁷