Skip to content
BeginnerEvidence: Grade Bsleep

The Deep Sleep Stack

A five-compound sleep protocol added one layer at a time, so you know which one is working. Layer 0 is the sleep problems a supplement cannot fix: apnea, a drifting clock, caffeine and alcohol. Layer 1 is glycine 3 g, the compound with the clearest human sleep data. Layer 2 is magnesium L-threonate, with a wearable trial behind it. Layer 3 is L-theanine for a racing mind. Layer 4 is apigenin, kept optional because the chamomile trial in insomnia was null on sleep diaries. Layer 5 is evening ashwagandha for people whose problem is cortisol rather than sleep itself. Each layer has a four-week test and an exit rule.

5 steps·5 compounds·Published May 8, 2026

Daily Schedule

Timing and dosage for each step

Layer 1 — 30–60 min before bed, from week 0

GlycineTier A

3 g

3 g dissolved in water. The clearest human sleep data of the stack: in sleep-restricted volunteers, 3 g before bed reduced next-day sleepiness and fatigue and improved reaction-time performance versus placebo. Run it alone for the first two weeks so you can see its effect.

Layer 2 — 1–2 h before bed, from week 2 if sleep onset or depth is still the problem

2 g

About 144 mg elemental magnesium. A 2024 randomized trial in adults with self-reported sleep problems improved wearable deep-sleep and REM scores versus placebo; a 2025 trial improved self-rated sleep impairment but not ring-measured sleep. Expect a modest effect, not a sedative one. Any magnesium form lowers the bar if the budget is tight: 500 mg of magnesium oxide shortened sleep onset and improved efficiency in elderly insomnia over eight weeks.

Layer 3 — 30–45 min before bed, from week 4 if the barrier is a racing mind

200 mg

200 mg. In a four-week crossover trial in 30 healthy adults, 200 mg/day lowered Pittsburgh Sleep Quality Index scores and anxiety ratings. It calms arousal; it does not sedate. No caffeine after midday while you test it.

Layer 4 — OPTIONAL, 30–60 min before bed with a little fat, from week 6

ApigeninTier A

50 mg

50 mg. The weakest layer: the only randomized trial of a standardized chamomile extract in chronic insomnia found no significant difference from placebo on sleep diaries over 28 days, with small effects on daytime functioning. Keep it if you notice a difference in your own diary; drop it without regret if you do not.

Layer 5 — with dinner or 1 h before bed, from week 0 ONLY if your problem is evening cortisol (wired-but-tired, waking 2–4 a.m., high stress)

600 mg

600 mg of a standardized root extract. A meta-analysis of five RCTs (400 participants) found a small but significant effect on sleep (SMD −0.59), larger at 600 mg/day or more, after eight weeks or more, and in people with insomnia. Serum cortisol fell significantly in a 60-day RCT. Not in pregnancy or with hyperthyroidism; cycle 8–12 weeks on, 4 off.

Read This Before the Protocol

Most sleep complaints that end up on a supplement page have a cause that no capsule reaches. Loud snoring, witnessed pauses in breathing or waking unrefreshed every day despite enough hours are apnea until a sleep study says otherwise. Falling asleep at 2 a.m. and waking at 10 a.m. refreshed is a clock problem, not a sleep-depth problem, and it has its own protocol: The Circadian Optimization Protocol. Caffeine after noon, alcohol in the evening and a bright phone at midnight each do more damage than any compound below can undo.

The second thing to know is what the trials actually measured. The glycine studies measured next-day sleepiness and performance in people whose sleep was cut short. The magnesium trials measured wearable scores and questionnaires, not polysomnography. The chamomile trial in diagnosed insomnia was null on the sleep diary. None of these compounds is a sedative, and the honest expectation is a modest improvement in how fast you fall asleep, how often you wake and how you feel in the morning.

The third: this is a stack of five, and you add them one at a time. The point of the layer order is that, four weeks in, you know which compound is earning its place. Taking all five on night one tells you nothing. Background and mechanisms are in The Deep Sleep Stack: Magnesium Threonate + Apigenin + L-Theanine + Glycine.

Layer 0 — Rule Out What Supplements Cannot Fix

CheckWhy it matters
Snoring, breathing pauses, morning headachesPossible sleep apnea; needs a sleep study, not glycine
Bedtime and wake time drift more than 1–2 hours between workdays and days offCircadian problem; go to the circadian protocol first
Caffeine after noon, alcohol within 3 hours of bedBoth fragment sleep; fix before testing anything
Night-time light, bedroom above 19–20 °CCheap, large effects
New medication, pain, reflux, nocturiaTreat the cause
Waking at 2–4 a.m. wired, high daytime stressThe cortisol pattern that points to Layer 5

Keep a two-week sleep diary before starting: bedtime, estimated time to fall asleep, number of wake-ups, wake time, how you feel at 10 a.m. on a 1–5 scale. Without a baseline, you will not be able to tell a 15-minute improvement from noise.

Who This Is For, and Who Should Not Run It

For: adults who fall asleep slowly, wake during the night or wake unrefreshed, who have ruled out the Layer 0 causes, and who want an alternative to sedative-hypnotics. Shift workers and people who are temporarily sleep-restricted fit the glycine data best.

Not for, without a clinician:

  • Pregnancy or breastfeeding — ashwagandha is excluded, and the rest lack pregnancy data.
  • Hyperthyroidism or thyroid medication — ashwagandha can push thyroid hormone up.
  • Kidney disease — magnesium clears through the kidney.
  • Benzodiazepines, Z-drugs or sedating antidepressants — do not add a stack on top of a prescription without the prescriber; apigenin acts at the same receptor site as benzodiazepines.
  • Suspected sleep apnea — a sleep study first. Sedation of any kind can worsen untreated apnea.
  • Children and adolescents — no data.

The Layer Logic

LayerWhatEvidence gradeWho
0 — Rule outApnea, clock, caffeine, alcohol, light—Everyone, two weeks of diary
1 — Glycine 3 gCore-temperature drop, faster onsetBEveryone, weeks 0–2 alone
2 — Magnesium L-threonateSleep depth, morning functionBFrom week 2 if onset or depth is still the problem
3 — L-theanine 200 mgPre-sleep arousalBFrom week 4 if the barrier is a racing mind
4 — Apigenin 50 mgGABA-A modulationC, optionalFrom week 6; keep only if your diary shows it
5 — Ashwagandha 600 mgEvening cortisolBFrom week 0 only for the cortisol pattern

The protocol is graded B. Its load-bearing layers are glycine and magnesium, each with randomized human data on sleep outcomes; the apigenin layer is grade C and optional, and the protocol does not depend on it.

Layer 1 — Glycine, Weeks 0–2 Alone

Dissolve 3 g in water 30–60 minutes before bed. It tastes sweet and mixes easily. In a placebo-controlled study of volunteers whose sleep was restricted by 25% for three nights, 3 g of glycine before bed reduced daytime sleepiness and fatigue and improved performance on a reaction-time task the next day; earlier work by the same group reported faster sleep onset and better subjective sleep quality. The mechanism is a drop in core body temperature through peripheral blood flow, the same signal the body uses to start sleep. Profile: Glycine.

What to expect: falling asleep faster is the first thing people notice, usually within the first week. If nothing changes in two weeks of a consistent bedtime, glycine is not your limiting factor; keep it anyway, it is cheap and safe, and move to Layer 2.

Layer 2 — Magnesium L-Threonate, From Week 2

Take about 2,000 mg of magnesium L-threonate, which delivers about 144 mg of elemental magnesium, one to two hours before bed. Split dosing (morning and evening) works too. In a 2024 randomized trial in adults with self-reported sleep problems, the threonate form improved wearable-measured deep-sleep and REM scores and several morning-function ratings versus placebo. A 2025 trial found better self-rated sleep impairment and cognition, but no difference on the sleep-tracking ring except in the subgroup with worse sleep. Read that as a modest, real effect that is easier to feel than to measure. Profile: Magnesium L-Threonate.

If budget matters, plain magnesium is a reasonable substitute for sleep purposes: 500 mg of magnesium oxide for eight weeks shortened sleep onset, improved sleep efficiency and lowered insomnia scores in elderly insomnia, though total sleep time did not change. Pooled with two other trials, the effect is about 17 minutes less time to fall asleep; the full trial record is in Magnesium for Sleep and Anxiety: What the Trials Show. The threonate form's case rests on brain penetration and the cognitive data, not on a head-to-head sleep trial.

What to expect: two to four weeks before you can judge it. Loose stools mean too much magnesium; reduce or split.

Layer 3 — L-Theanine, From Week 4

Add 200 mg 30–45 minutes before bed if your diary says the problem is a mind that will not switch off rather than a body that will not settle. In a four-week randomized crossover trial in 30 healthy adults, 200 mg/day lowered sleep-quality-index scores and trait anxiety ratings. It raises alpha-wave activity and takes the edge off arousal without sedation, which is why it pairs with the first two layers rather than replacing them. Profile: L-Theanine.

Do not evaluate L-theanine while drinking coffee after midday. You will be testing caffeine, not theanine.

Layer 4 — Apigenin, Optional, From Week 6

Apigenin, the main flavonoid in chamomile, binds the benzodiazepine site of the GABA-A receptor. That is a good mechanistic story, and the trial record is thin: the one randomized trial of a standardized chamomile extract in chronic primary insomnia (28 days) found no significant difference from placebo on total sleep time, sleep latency or awakenings, with modest, non-significant advantages on daytime functioning. The 50 mg dose on this page is a community convention, not a trial dose. Profile: Apigenin.

Keep it only if your diary shows a difference after two weeks. This is the layer to drop first.

Layer 5 — Ashwagandha, for the Cortisol Pattern

This layer is not for everyone. It is for the person who is wired at bedtime, wakes at 2–4 a.m. with a racing heart, and carries high daytime stress: the pattern that points to evening cortisol rather than sleep mechanics. If that is you, 600 mg of a standardized root extract with dinner or an hour before bed, from week 0, alongside glycine. A meta-analysis of five randomized trials (400 participants) found a small but significant improvement in sleep (SMD −0.59), with larger effects at 600 mg/day or more, after at least eight weeks, and in people with diagnosed insomnia. In a 60-day RCT of a high-concentration root extract, serum cortisol fell significantly versus placebo, and a 2019 trial at 250 and 600 mg/day reduced perceived stress and cortisol and improved sleep quality. Profile: Ashwagandha; the stress side is covered in Managing Cortisol: Ashwagandha, Selank and Magnesium.

Cycle it: 8–12 weeks on, 4 weeks off. Not in pregnancy, not with hyperthyroidism or thyroid medication, and stop if you notice unusual drowsiness, stomach upset or jaundice.

What Is Deliberately Left Out

  • Melatonin. It is a clock signal, not a sleep-depth compound, and it belongs to the circadian protocol, where timing is the whole point.
  • DSIP and other research peptides. No human dosing trial supports a dose. Background in Deep Sleep Protocols: DSIP, Glycine and Sleep Architecture.
  • Antihistamines and over-the-counter sleep aids. They fragment sleep architecture and build tolerance.
  • Alcohol as a sleep aid. It shortens sleep onset and then fragments the second half of the night.

Monitoring

MarkerHowWhenWhat counts as working
Time to fall asleepDiary estimateNightly15 minutes or more faster than the two-week baseline
Night wake-upsDiary countNightlyFewer than baseline across a week
Morning feel, 1–5Diary at 10 a.m.DailyA full point better across a week
Deep sleep and HRVWearable, if you have oneWeekly averageDirection, not single nights; wearables are noisy on sleep staging
Side effectsLoose stools, grogginess, vivid dreamsAs they occurAdjust the layer that was added last

Judge each layer on weekly averages against the baseline diary, never on a single night.

Exit Criteria

  • If two layers in a row change nothing over their four-week windows, stop adding. The problem is not what this stack treats; go back to Layer 0 and see a clinician about apnea, mood or pain.
  • If snoring, breathing pauses or daytime sleepiness while driving appear at any point, stop and get a sleep study.
  • If you are still relying on the stack at six months and sleep collapses without it, that is a sign to look for the cause rather than to keep stacking.
  • Taper nothing: none of these compounds needs tapering. Stop the ones your diary does not justify.

Related Research