Pharmacists who use Claude quickly discover the same frustration. You ask it to reason through a drug interaction or help you phrase a counselling point for a complex polypharmacy patient — and the answer is generic. Claude doesn't know you're a clinical specialist, doesn't know your formulary, doesn't know the protocols your service follows, and doesn't know you already asked a related question last Thursday and landed on a different approach than the textbook default.
Every session starts from nothing. That's the problem Stash solves.
The amnesia is expensive in pharmacy because the context that makes Claude useful is highly specific:
Without persistent context, Claude gives you UpToDate answers when you need answers calibrated to your actual practice.
Stash stores your professional context — not patient data. The distinction matters and it's intentional. What goes in Stash:
| Collection | What goes in it |
|---|---|
formulary | Preferred agents, local choices, rationale for deviations from national guidance |
protocols | Your service's specific monitoring intervals, dose adjustments, escalation thresholds |
counselling | Phrases, analogies, and explanations that have worked well in patient-facing conversations |
interactions | Notes on combinations you've researched and the clinical conclusion you landed on |
context | Your specialty, patient population, standing preferences, and how you like Claude to respond |
No patient names, no dates of birth, no identifiers. Your practice knowledge — the professional layer that sits above any individual case.
You: "context()"
Claude: [loads your specialty context: clinical pharmacy, anticoagulation service, your formulary notes, your preferred monitoring intervals]
You: "I need to explain warfarin-clarithromycin interaction risk to a patient who's been stable on warfarin for 18 months. They need 5-day macrolide course. What's the core message and what do I tell them about INR checks?"
Claude: "Based on your anticoagulation context and the 3-day monitoring interval your service uses for significant interactors: the core message is that clarithromycin inhibits warfarin metabolism — their INR can rise meaningfully within 48–72 hours even on a stable dose. I'd frame it as: 'Your warfarin is working well but this antibiotic changes how your body breaks it down. We need an INR check in 3 days — that's not optional, it's part of the course.' Your formulary notes show amoxicillin as the preferred alternative where clinically appropriate — worth confirming whether spectrum matters here."
That answer knew your monitoring interval, your formulary preference, and how to frame things for a patient conversation — because you stored that context once.
You: "search protocols for vancomycin monitoring"
Claude: [retrieves your stored protocol] "Your service uses AUC-guided dosing with a target of 400–600 mg·h/L. Initial trough-only approach is acceptable for short courses under 72 hours per your protocol note from March. Your note flags that the renal team prefers dose-hold at CrCl below 20 rather than extended interval — that differs from the national guidance."
That's your protocol, your team's deviation, available in ten seconds without hunting for the PDF.
One of the most useful patterns for clinical pharmacists: logging the interaction combinations you research, and the clinical conclusion you reached. Not for legal documentation — for your own working knowledge.
You: "add to interactions: 'Metformin + IV contrast — our radiology team pauses metformin 48h pre-contrast for eGFR below 45, resumes 48h post if stable. Based on MHRA 2023 update. Lower risk than historical practice suggested but not eliminated.'"
Claude: [saves to Stash] "Saved. Seventeen interaction notes in your library now."
Next time that question comes up — from a junior doctor, a nurse, or your own memory — you have the answer with the reasoning behind it, not a generic response from a clinical database that doesn't know how your unit decided to handle it.
Experienced pharmacists develop a bank of phrases that work — explanations that land with patients who are confused, anxious, or non-adherent. Claude can help develop and refine these, and Stash stores the ones worth keeping.
You: "add to counselling: 'For explaining why statins need to be taken even when cholesterol feels fine: use the boiler analogy — the boiler is working quietly in the background; you don't wait for a breakdown to service it. Landed well with patients who question long-term preventive therapy.'"
Claude: [saves] "Added. Six counselling notes now. Want me to search for your existing notes on preventive therapy framing?"
Stash stores professional context. It is not a clinical decision support system, it does not replace your BNF or Micromedex, and it does not store or process any patient information. The records you store are your practice notes — equivalent to the mental models and shorthand you carry as a clinician, made searchable.
Claude with Stash can help you draft counselling points, think through interactions, check your recall of a protocol, or frame a recommendation — it cannot substitute for clinical judgment, and nothing in Stash changes that. The stored context makes Claude's responses more calibrated to your practice; it does not make them more authoritative.
add to context: [your specialty, patient population, how you want Claude to respond]context() loads everything. Start working.Add Stash to Claude
Connector URL (Claude Settings → Integrations → Add MCP):
https://app.stashlite.com/mcp
Free tier. Google sign-in. No card required. Pricing may change; cancel anytime.