Loop, iAPS, Trio, and AndroidAPS — the community-built automated insulin delivery systems that preceded every commercial AID algorithm on the market. What they are, how they compare, and the biggest library of real-world Loop content anywhere.
Device specifications change frequently — always verify current information directly with the manufacturer before making any decisions. Full disclaimer.
Curious about DIY Loop? Start here.
This is not a commercial product. DIY AID systems are open-source software built and maintained by the diabetes community. They are not FDA-cleared (with some exceptions), not from a company, and not covered by a warranty or support line. You build and maintain it yourself — with community help. This is a choice made by informed adults and caregivers, not a default recommendation.
Commercial AID systems like Omnipod 5, Control-IQ, and the iLet took years to reach FDA approval. While they were waiting, people with T1D built their own. Starting around 2015, the community developed open-source algorithms that connected existing CGMs and pumps to create automated insulin delivery — years before commercial options existed. The results were remarkable enough that Tidepool eventually brought the Loop algorithm through the FDA, and it now powers the commercial twiist pump. The DIY ecosystem that exists today is larger, more customizable, and in some cases more advanced than its commercial descendants.
Scott has been covering DIY Loop since Arden started using it. She went through Loop, then iAPS, and now uses Trio — a next-generation fork of iAPS. The Juicebox Algorithm Pumping series has 39 episodes covering Loop, iAPS, Omnipod 5, Medtronic 780G, Control-IQ, and more. juiceboxpodcast.com → Series → Algorithm Pumping
The algorithms are free and open-source. You pay for the hardware (pump, CGM), but not the algorithm. For people with compatible existing hardware, the cost to start looping can be zero beyond time.
You install the app yourself — either by compiling from code or using community browser-based builders. There's no customer service phone number. The tradeoff for this complexity is unlimited customizability and often better performance for experienced users.
Support comes from Loop and Learn (Facebook), Looped (Facebook), LoopDocs documentation, and thousands of community members. The knowledge base is enormous — and it's people who actually live with T1D answering your questions.
DIY systems offer settings commercial systems don't allow — from super micro boluses (SMBs) to autotune, autosens, and dynamic ISF. Experienced users routinely achieve results that exceed commercial system benchmarks.
The original. Built for iPhone. Works with Omnipod EROS/Dash or older Medtronic pumps. Connects to Dexcom CGM. The foundation that twiist is built on commercially. Good starting point — cleaner, more straightforward than iAPS.
Builds on Loop's foundation with additional algorithm features from oref1 (originally from OpenAPS). Includes super micro boluses (SMBs), autotune, dynamic ISF. Significantly more powerful — and more complex. This is what Arden uses.
The Android equivalent. Works with a wider range of insulin pumps than Loop. More complex setup. Runs the oref1 algorithm (same as OpenAPS and iAPS). Good option for Android users or those with non-Omnipod pumps.
Using the algorithm and watching the data come back from it and seeing how it reacts will absolutely supercharge your understanding of how insulin is working in your body. I'm telling you — watch an app, a computer algorithm decide about insulin. It just elevates your understanding.
— Scott Benner · Juicebox Podcast Episode #313 · Fox in the Loop House Part 2A meaningful percentage of your tech-engaged T1D patients are running DIY AID — and many don't tell their clinician. Ask. Looping patients often have excellent glycemic control and can explain their settings in detail.
DIY systems use standard pump settings — basal rates, I:C ratio, ISF, DIA (duration of insulin action). Your clinical input on these settings is as important as with any commercial system. Bad settings → bad results, regardless of algorithm.
DIY systems upload to Nightscout (open-source remote monitoring), Tidepool, or both. Many looping patients can share real-time data with caregivers and clinicians. Ask what platform they're using before assuming there's no data.
You don't need to understand the DIY software to support a looping patient. Your role is to help optimize their underlying pump settings and ensure their CGM is accurate. The algorithm handles the rest — often better than commercial alternatives.
DIY AID requires three things: a compatible insulin pump, a compatible CGM, and the software. The software is free. The rest is what makes it complicated — hardware compatibility has changed a lot over the years, and the system you choose determines what hardware you need.
Compatible Hardware — 2025/2026The DIY AID systems aren't just experiments — they have published clinical evidence, rigorous community study, and in some cases match or exceed commercial system outcomes. Here's the full picture, including how Loop evolved into what powers the commercial twiist pump.
Algorithm ArchitectureLoop's core algorithm predicts future glucose using insulin on board and CGM trend. It adjusts basal delivery every 5 minutes — increasing to prevent highs, decreasing or suspending to prevent lows. No automatic boluses beyond basal adjustment. Clean, effective, and the origin of Tidepool Loop (now in twiist).
iAPS/Trio and AndroidAPS add the oref1 algorithm layer, which introduces Super Micro Boluses (SMBs) — small automatic correction boluses delivered every 5 minutes when glucose is trending up. This is more aggressive than pure basal adjustment and tends to produce better post-meal outcomes for users who have good settings dialed in.
Available in iAPS/Trio and AndroidAPS — an automated analysis that reviews 7 days of data and recommends adjustments to your basal rates, I:C ratio, and ISF. Removes the guesswork from settings optimization. Not available in commercial AID systems.
Autosens detects insulin sensitivity changes over time and adjusts algorithm aggressiveness accordingly — useful for illness, exercise recovery, hormonal cycles, and growth spurts. Dynamic ISF extends this further. Commercial systems have their own sensitivity adaptations but typically less transparency and customizability than open-source versions.
Watching the Loop take away basal, give it back — it was like a self-driving car. It was seeing the road curve a little down or a little up and just going with it. And it felt like it had a crystal ball. Blood sugar goes from 85 to 89 and all of a sudden her basal goes up a little. And you think — it's not even a rise yet.
— Scott Benner · Juicebox Podcast Episode #420 · Fox in the Loop House Part 3DIY looping patients who have been using these systems for years typically have deep knowledge of insulin pharmacokinetics, pump settings, and algorithm behavior. They are often your most technically sophisticated T1D patients. Listen to them.
Ask for Nightscout access or Tidepool data. DIY systems often produce richer data than commercial systems — including detailed logs of every algorithm decision. Some patients will show you displays most commercial software doesn't offer.
Autotune helps, but your input on basal rates, I:C ratio, and ISF remains clinically important. Bad settings + powerful algorithm = worse outcomes. Good settings + powerful algorithm = excellent outcomes. This is still a partnership.
Standard open-source Loop/iAPS/AndroidAPS are not FDA-cleared (Tidepool Loop is). Prescribing or recommending these systems puts clinicians in legally ambiguous territory. Many clinicians choose to support — not prescribe — DIY patients. ADCES has published guidance on supporting DIY AID users.