OpenAPS (oref0)

OpenAPS — the Open Source Artificial Pancreas System — is the project that started DIY closed-looping. After building a personal system called #DIYPS, Dana Lewis and Scott Leibrand (with Ben West, who reverse-engineered communication with older Medtronic pumps) open-sourced their work as OpenAPS in February 2015, under the community banner #WeAreNotWaiting.¹ Its purpose is "an open and transparent effort to make safe and effective basic Artificial Pancreas System (APS) technology widely available" to anyone willing to build their own.²

Platform. Classic OpenAPS is not a phone app. It runs on a small Linux "rig" — historically an Intel Edison + Explorer Board, or a Raspberry Pi — that sits between your pump and continuous glucose monitor (CGM) and adjusts insulin every five minutes.¹ The Edison hardware is discontinued, which is a large part of why the project is now legacy-leaning.³

Algorithm lineage and sophistication. OpenAPS publishes the reference design and the oref0 / oref1 algorithm that much of the field builds on.⁴ It calculates insulin-on-board, predicts where glucose is heading, and sets temporary basal rates toward a target.⁴ oref1 added super-micro-boluses (SMB) for faster mealtime response and unannounced-meal (UAM) handling, plus autosensitivity that adapts to changing insulin needs.⁴ This code lineage powers AndroidAPS and Trio (with adaptations), so OpenAPS's importance far exceeds its current direct user base.⁵

Supported pumps and CGMs. Direct OpenAPS works only with specific older Medtronic MiniMed pumps (e.g. 515/715, 522/722, and 554/754 Veo on early firmware), because newer firmware blocks the remote commands looping needs.⁶ Supported CGMs include Dexcom G4/G5/G6 and Medtronic Enlite, with other sensors usable via Nightscout.⁷

Customizability. Because it is the open reference implementation, parameters and dosing decisions are inspectable and tunable in open source — the core promise of the movement.⁴

Community, docs, setup and legal status. Documentation is extensive (the OpenAPS ReadTheDocs), but building a rig is genuinely hard: flashing a single-board computer and maintaining it from the command line.¹ Crucially, OpenAPS is not a regulated or commercial product — there is no company behind it, and the trademark is "not authorized for use by any commercial entity"; you build it and you are responsible for it.²

Clinical evidence. This is where OpenAPS is unusually strong for a DIY tool. The CREATE randomized controlled trial tested the OpenAPS 0.7.0 algorithm (inside a modified AndroidAPS) against a sensor-augmented pump in 97 children and adults; time-in-range rose from 61% to 71% with the loop versus a fall in controls — about 3 hours 21 minutes more per day in range — with no severe hypoglycemia or ketoacidosis.⁵ Its 24-week continuation phase (about 48 weeks total use) confirmed a sustained ~12-point time-in-range benefit and a 0.5% HbA1c reduction.⁸ Observational and systematic-review data — including a pediatric survey and a systematic review of 730 participants — consistently show improved time-in-range, lower HbA1c and good quality of life, while noting most pre-RCT evidence was observational.⁹¹⁰

What's coming. The future of this algorithm is largely its descendants: oref code lives on and evolves inside AndroidAPS and the newer Trio, where active development now concentrates.⁵³ OpenAPS itself remains available and its reference design remains published, but the oref0 codebase has had no new release since 2022.³

The algorithm also remains a research substrate. A 2026 research letter in Diabetes Care from a Stanford group reports that pairing high-potency GLP-1 receptor agonists with the OpenAPS automated insulin dosing algorithm removed the need to announce meals in order to hit glycemic goals.¹¹ Two caveats worth holding onto: this is a short letter rather than a full trial report — it carries no published abstract, so we can point to its conclusion but not to effect sizes, sample size or how durable the result is — and GLP-1 receptor agonists are not approved for type 1 diabetes. Treat it as an early and interesting signal about where unannounced-meal looping might go, not as a recipe to copy.