Aminoacid documentation

What this is

Aminoacid is a constrained LLM that takes a peptide stack, goal, experience tier, and optional bloodwork notes, and returns a sequenced 4–12 week cycle: per-compound dose, injection cadence, washout windows, bloodwork checkpoints, and red-flag thresholds.

The agent does not prescribe. It composes a literature-aware draft from published kinetic data and protocol-community consensus. Every output ends with a clinician-review disclaimer because that is what every output requires.

Peptide classes the agent recognises

Each peptide is classified by its primary receptor target. The agent uses class membership to detect synergy, receptor competition, and cross-class redundancy.

Growth-hormone axis

Regenerative / repair

Melanocortin system

Other

• Selank, Semax — nootropic / anxiolytic; nasal route; no chronic safety data > 12 weeks
• Epitalon (epithalon) — putative pineal/telomere effects; evidence sparse
• DSIP — sleep-related peptide; mostly anecdotal data

Pharmacokinetics

Plasma half-life drives injection cadence. The agent never picks "twice a week" because someone said so — it's derived from each compound's measured t½.

Stack interactions

Pairwise rules the agent applies before composing a cycle:

• GHRH + GHS = synergy. Pulsatile GHRH analog (cjc-1295 no DAC) plus a GHS (ipamorelin) produces supra-additive GH amplitude. This is the standard "fragment-pair" stack.
• Two GHRH analogs = competition. cjc-1295 + tesamorelin (or + sermorelin) both bind GHRH-R. Stacking yields no extra benefit and wastes compound. Flagged as avoid.
• Chronic GHS = receptor desensitisation. Hexarelin in particular; ipamorelin tolerates longer use. Cycles cap GHS exposure to 8–12 weeks with washout.
• BPC-157 + TB-500 = parallel pathways. Cytoprotective + actin-binding don't compete. Often co-cycled for injury recovery.
• MT-II + PT-141 = MC-receptor overlap. Both hit MC4R. Combined dosing risks cardiovascular stacking; flagged as caution.
• MK-677 + GH peptides = redundant signal. Both elevate GH/IGF-1; combined raises fasting-glucose risk. Flagged for bloodwork follow-up.

Bloodwork markers by class

The agent picks marker panels from the stack composition, not from a generic checklist.

Washout windows & cycle length

• GH axis — 8–12 weeks on, 4 weeks off. Receptor desensitisation accumulates after ~12 weeks of continuous GHS exposure.
• Regenerative (BPC-157, TB-500) — 4–6 weeks on, 2 weeks off. Often run shorter targeted blocks for specific injuries.
• Melanocortin tanning (MT-II) — load phase 2–3 weeks, then maintenance 2×/week. Stop if BP drift exceeds threshold.
• MK-677 — limit to 8–12 weeks. Edema and glucose drift trend up by week 6.

The agent

Inference. The agent runs a research-tuned language model with a strict educational-simulation system prompt. Output streams in roughly ten seconds.

Determinism. Same inputs ≠ identical output by design. Minor variation in dose ranges prevents "the AI told me exactly 250mcg" copy-paste. Ranges are returned, not point values.

System prompt design

The agent's behaviour is shaped by a hard-coded system prompt that:

• Frames the output as a research literature simulation, never as a prescription.
• Forces dose ranges, not point doses, calibrated by user-declared experience tier.
• Refuses to recommend vendors, sources, or compounds outside the published research-peptide canon.
• Enforces output structure: weekly schedule, washout, bloodwork list, red-flag thresholds, clinician-review footer.
• Adds explicit caveats when the stack contains a compound with sparse human data (epitalon, DSIP).

You can read the live prompt in the repo at api/_lib/prompt.ts.

Data flow

browser ──► /api/plan-cycle ──► AI agent
                  │                   │
                  ├── auth (Privy)    └── plan text
                  └── input ─ stack, goal, experience, optional bloodwork notes

returned plan ──► browser (rendered inline)
                  │
                  └── metadata only (date, goal, # peptides, experience tier)
                       saved to Privy custom-metadata historyJson · max 10 entries

payment    ──► /api/solana-rpc ──► Helius RPC ──► Solana mainnet
                                         │
                                         └── tx signature ──► /api/verify-payment
                                                                    │
                                                                    └── Privy custom metadata
                                                                       { plan: 'pro', paidSig, paidAt }

What we never persist: bloodwork notes, plan text, IP addresses, user-agent, page-view analytics. The only thing we store is the cycle counter and Pro-status flag in Privy custom metadata.

Anonymous tier: a HMAC-signed cookie tracks 1 free anonymous cycle per browser. Trivially cleared in incognito; intentional soft-gate.

Limits & failure modes

• The agent does not measure your blood. If you input "IGF-1 high-normal" we trust you. Garbage in, garbage out.
• Receptor competition rules are pairwise. Three-way interactions in exotic stacks aren't covered comprehensively.
• Compound purity isn't checked. Reconstitution math, vendor reliability, and cold-chain handling are out of scope.
• Long-term safety beyond ~12 weeks per cycle is not well studied for most research peptides. The agent caps cycle length to published windows; chronic use isn't endorsed.
• The agent declines on AAS/SARM combinations, anti-aging "blueprints," and any input attempting to elicit prescription-grade dosing.

References

• Veldhuis JD, et al. Pulsatile growth hormone secretion in humans: a brief review. Endocr Rev.
• Sigalos JT, Pastuszak AW. The safety and efficacy of growth hormone secretagogues. Sex Med Rev. 2018.
• Chang CH, et al. BPC 157 and gut–brain axis. Curr Pharm Des.
• Goldstein AL, Hannappel E, Sosne G, Kleinman HK. Thymosin β4: a multi-functional regenerative peptide. Expert Opin Biol Ther.
• Diamond LE, et al. Bremelanotide for hypoactive sexual desire disorder (RECONNECT). Obstet Gynecol. 2019.
• Falutz J, et al. Effects of tesamorelin on visceral fat in HIV. NEJM. 2007.

Citations are illustrative — not exhaustive. Always cross-check the underlying literature before acting on any cycle.