PillScout
Know what everything you take is doing.
Claude can answer questions. It can't track everything you take, verify it against curated safety sources with citations, or see how risk builds across your full cabinet. That's what PillScout is for. It remembers what you take, checks it against verified data, and flags interactions and cumulative patterns with clear sourcing. You see exactly what's backed by evidence and what isn't. No guessing, no overreach, just the information as it stands.
Request a beta keyPillScout is in closed beta.
What's in the engine
PillScout runs as an MCP server that connects to Claude. When you ask about something you take, it checks your full cabinet instead of answering from memory.
Persistent cabinet. Connect once. Your full list travels across every Claude conversation. No re-listing what you take every time you ask.
Verified ingredient profiles. 150+ profiles with citations covering the medications, supplements, and compounds people actually take.
Real-world safety checks. Evaluated across your full cabinet, not one item at a time.
Interactions. Drug–drug, drug–supplement, and supplement–supplement interactions, with mechanism where known.
Cumulative burden scoring. Anticholinergic load, CNS depression, serotonin activity, QTc prolongation risk, and other additive effects.
Stacking detection. Patterns like triple whammy combinations, hepatotoxic stacks, and overlapping mechanisms that increase risk together.
Age-specific rules. Beers Criteria for adults 65+, with AGS 2023 citations.
Pregnancy and lactation. Checks against FDA PLLR, Briggs, and LactMed guidance.
Condition-specific flags. Contraindications and cautions tied to known conditions in your context.
Symptom-to-cabinet attribution. Maps reported symptoms back to plausible contributors in your cabinet, with confidence levels.
Timing and separation. Absorption conflicts, spacing requirements, and sequencing considerations.
CYP pathways. Metabolic interactions that affect exposure levels.
Bleeding risk. Additive anticoagulant or antiplatelet effects across combinations.
Every flag includes an applicability note in plain language: what triggered it, when it matters, and what it does and doesn't imply. It's information, not a verdict.
Patient context. Age, conditions, pregnancy, and lactation status travel with your cabinet. Rules apply when they should, and don't when they shouldn't.
How it works. Claude retrieves. PillScout verifies, scores, and cites. When a question touches your cabinet, Claude pulls the context. PillScout runs the check.
Each result is built on three things:
Verification. Ingredient-level checks against curated sources, with citations attached to every finding.
Scoring. Interactions and cumulative burden evaluated across your full cabinet, not in isolation.
Clarity. Every output states what the evidence supports, what it suggests, and what it doesn't establish.
This isn't a single lookup. It's a pass across everything you take, with cumulative risk treated as first-class, not an afterthought. Every flag comes with a source and an explicit note on what that source does and doesn't tell you.
Coverage signals: what's verified, what isn't
Every safety engine has gaps. The question is whether it tells you where they are.
PillScout shows you what's database-verified, what's inferred from class rules, and what's pending label verification.
Gaps aren't failures. They're transparency. A clean result with full coverage means something. A clean result with partial coverage means something different. You get to see which one you have.
How PillScout gets better
Every time someone uses it, the next gap becomes visible.
When an ingredient or combination isn't fully covered, that missing piece is recorded along with how often it comes up. Not your cabinet. Not your identity. Not your conversations. Only the structured inputs to the safety tools.
That means the database grows based on what people actually ask about. The most common missing pieces become the next verified profiles.
Part of a bigger picture
You can already track a lot.
Steps, heart rate, and daily activity with Apple Health. Deeper sleep and recovery data with Oura Ring. Nutrition with MyFitnessPal.
Each one tells your AI something real about your life.
PillScout is the one that knows what you take. And without that, the picture is incomplete.
If your AI sees your sleep but not your magnesium, your heart rate but not your medication, your nutrition but not your supplements, it's missing a key part of what's actually driving how you feel.
Once your cabinet is there, your AI can reason across everything:
"I slept poorly last night. I took magnesium. Did it help?"
"Is this supplement affecting my heart rate?"
Those become questions your assistant can actually answer.
I'm ScoutI don't guess what you takeI check it
How PillScout got here
I started PillScout because the tools I wanted didn't exist, or took too much work to be useful.
I'd see a supplement making big claims and wonder: can I take this with what I'm already on? A family member is on heart medication, would this help or hurt them? My mom takes prescriptions from different doctors and a bunch of supplements and always feels nauseous. Is one of them causing it? Could another one actually help?
Checking meant searching, piecing things together, and usually giving up.
AI can answer now, but only if it knows everything you take.
I asked my kids to build it. They said, "you build it." So I did.
I started with a web app because that's what was possible. Then MCP came along, and the version I actually wanted became possible: a cabinet your AI assistant uses automatically, wherever you're already talking to it.
That's what PillScout was always meant to be.
Request a beta key
PillScout is in closed beta. I'm handing keys out one at a time to clinicians, researchers, builders, students, caregivers, and anyone curious about health tech.
Email pillscout@pillscout.health with a line about who you are and what you'd try it on. I read every one.
If you'd also like to see the beta web app, mention it in your note.