Psychiatry.Ink is a clinical writing workspace built specifically for psychiatric documentation: admission histories, progress notes, psychopathological findings, diagnoses, medication plans, labs, and therapy planning — all in one calm, case-centred workspace, with optional AI assistance that always waits for your review.

After signing in you land on the dashboard ("My patients"). Click the new-patient button, enter a name or working label, and the case opens with all clinical areas: history, progress notes, diagnosis, labs, medication, and therapy. Patient identifiers are encrypted in your browser before anything is stored.

Yes. On first login a short 4–5 step tour walks you through the workflow, the clinical areas and the role of the AI. It is skippable and can be restarted any time from the Help menu. The dashboard shows a "Getting started" checklist that guides you to your first finished document, small first-use hints explain the key controls once, and empty sections tell you what to do next. Subtle suggestions — computed entirely on your device — help you structure notes, and nothing is committed without your approval.

The full interface is available in German, English, French, and Spanish. The public site language follows the domain (psychiatry.ink is English, psychiatrie.ink is German); inside the app you can pin your preferred language in the settings at any time.

No. Psychiatry.Ink runs entirely in a modern browser on desktops and tablets — there is nothing to install and nothing to maintain. The workspace is optimised for larger screens, since clinical documentation needs room.

Yes. A synthetic demo patient is provisioned for new accounts so you can explore documentation, diagnosis support, medication plans, and lab trends without touching real data. All demo material is clearly marked and contains no real patient information.

Patient identifiers (name, date of birth) are encrypted in your browser and stay encrypted at rest, with the keys held on your device. Clinical content is stored end-to-end encrypted on EU-hosted infrastructure; we never persist or log plaintext identifiers. For some AI and dictation features, identifiers may be transmitted transiently to the server so clinical text can be de-identified or transcribed before it reaches a model — used only for that processing, never stored and never logged. An optional account backup, encrypted with your passphrase, is available for recovery and moving between devices.

Identifiers do not. Two independent layers protect every AI call: identifiers are encrypted client-side and are never part of AI requests, and the server additionally de-identifies all clinical text before it reaches any model. If residual identifying patterns are detected, the request is blocked instead of sent.

No — by design. Scans and PDFs can carry patient identifiers in the image itself (letterheads, stamps) that cannot be reliably removed, so the system does not read them at all: the file is stored encrypted as a document you can view, and no content is extracted (no OCR, on the device or the server). Structured data from a scan is entered manually. Only text formats (DOCX, TXT, CSV, XLSX, JSON) are parsed automatically.

Each case carries only a random identifier (a UUID). The link between that identifier and the real patient — name and date of birth — exists only as ciphertext: on your device in the vault, and, if you use the encrypted account backup, on the server only inside a block encrypted with your passphrase. The server can see, for example, that "case a1b2c3… has an encrypted 12 KB snapshot," but not that a1b2c3… is "Jane Doe, born 1980." Even if our servers were breached, only ciphertext and non-identifying metadata would be exposed — no names, no mapping, no case contents, and no scanned documents (those never leave your device).

The keys that protect patient identifiers exist only on your devices — we cannot read them, and neither could anyone who compromised our servers. The flip side: you are responsible for your device access and backups. The app offers an encrypted account backup with a passphrase for recovery and device transfer.

Four modes, selectable under Settings → Privacy: "Local only" (everything stays in this browser), "Sync patient list" (list available across devices, case files local), "Sync everything encrypted", and "Sync case file only" (clinical content end-to-end encrypted in the cloud while name and date of birth stay on the device). The app requests persistent browser storage at sign-in so the browser does not evict local data automatically. Still important: manually clearing browser or site data irrecoverably deletes all local cases — export a vault backup regularly or enable one of the sync modes. In Chrome and Edge you can additionally pick a local backup folder: the app automatically stores encrypted backups of your case files there that survive even clearing browser data — restorable with your passphrase.

Only if you share them. Team workspaces support roles and per-case access with audit logging; case discussions with colleagues use de-identified case packages by default. Nothing is shared automatically.

The product is designed around GDPR principles: data minimisation, client-side encryption of identifiers, EU hosting, de-identification before AI processing, and full export of your data at any time. A data processing agreement (DPA) and the current list of sub-processors are published on this site.

The Architecture page explains the core design in plain language: the zero-knowledge patient vault, the single guarded gateway every AI call must pass, and the two named systems behind the workspace — Mira, the knowledge curator, and Noevia, a deterministic clinical inference engine now in beta. The Security page adds step-by-step diagrams of the encryption with concrete examples.

Drafting and structuring clinical text, dictation with transcription, inline voice editing of selected text, diagnostic criteria support, summaries, and medication-related checks. Every AI output is a draft: nothing is written into the record until you review and accept it.

No. Diagnostic criteria support (Butterfly) evaluates documented criteria deterministically first and uses AI only for unresolved points — and even then the output is advisory with evidence quotes. A diagnosis only enters the record through your explicit clinical decision and attestation.

NOEVIA is a deterministic, rule-based analysis that turns a case's documented findings into a structured clinical overview — dimensional profiles and mechanism hypotheses across 27 dimensions and 15 mechanism domains, each traceable back to the source documentation. It runs entirely on your device, uses no language model in its scoring and consumes no AI credits. It is advisory only: it never makes a diagnosis and never writes to the record — your clinical judgement always decides. NOEVIA is now in beta with a limited scope (currently a few dimensions and mechanisms carry active rules; the rest show "not evaluated"), and its scoring weights are drafts still pending clinical sign-off. You will find it as the NOEVIA tab inside a patient case.

AI actions (drafting, dictation, analysis) consume credits depending on the length and depth of the task. Tools like the short progress-note summary and Translate show the estimated cost before you start. Your plan includes a monthly credit allowance, and additional credits can be purchased. If your balance falls below 100 credits (and again below 25), you receive a one-time email warning — no repeated reminders. When credits run out, all writing, editing, and exporting keeps working — only AI actions pause.

Psychiatry.Ink routes AI tasks to leading models in three quality tiers — Economic, Standard, and Thorough — plus an explicit opt-in maximum tier for the hardest tasks. An EU-resident model option is available for practices that require it. All calls go through the same de-identification gateway regardless of provider.

The psychopharmacology knowledge base is curated with the help of Mira, our internal knowledge-curation system. Mira audits the database for gaps and contradictions, monitors sources such as PubMed, Europe PMC, EMA and FDA, and prepares fully traceable improvement proposals — and every risk-relevant change is reviewed by clinicians before it is published. Mira works exclusively on the knowledge base and has no access to patient data. The Architecture page describes how she works.

Yes. AI assistance is optional and per-action — nothing runs in the background on your notes without you triggering it. The workspace is fully functional as a pure documentation tool.

Admission and history documentation, longitudinal progress notes, psychopathological findings, diagnoses with ICD-10/11 coding, medication plans, lab documentation with trends, psychotherapy and complementary therapy planning, discharge summaries and letters, plus reusable document templates.

Click the dictation control in the editor, speak, and the transcription is inserted as text. Before the first recording you confirm via checkbox that the patient has consented; no recording starts without this confirmation (switching to another case asks again). You can also select any passage and give a spoken or typed instruction (keyboard shortcut ⌘⌥B) — the AI proposes a rewrite that you accept, reject, or rerun.

Yes. The detail view of a document in the Documents area offers a "Sign & lock" action. It records who signed and when; the document then becomes read-only — editing and "Open in workspace" disappear. Signed documents can still be duplicated, exported, and printed. To make changes, use "Duplicate" to create an editable, unsigned copy; the signed original stays untouched.

Yes. Open Visit mode from the dashboard: appointments from today's calendar are included automatically, further patients are added by search, and the order is arranged by drag & drop. Per patient, the visit view shows diagnoses, current medication (editable, incl. the last change), safety signals with an update-risk action, the latest progress entries, abnormal lab values and the last ECG — plus the same adaptive quick actions as the overview (guided progress note, dictation, risk update, medication change and more), a to-do box (tasks store only the case ID, never the patient name) and a quick-entry box for the progress note, optionally polished by AI (the suggestion is only applied after your review). Previous/next buttons move between patients, and the full case file is one click away.

Butterfly checks your documentation against operationalised diagnostic criteria, shows which criteria are met, unmet, or unclear, and suggests targeted questions for the unclear ones. You resolve each point clinically (present / absent / unclear), and your attestations — not the AI — are the source of truth.

A dedicated psychotherapy workspace with compact clinical tabs: overview, case formulation (guided SORKC behavioral analysis and Beck-style cognitive conceptualization plus macro analysis), therapy goals with success criteria, phase-based treatment planning, session documentation, exercises and homework, evaluation with score series and goal-attainment ratings, and relapse prevention. Sessions can be documented briefly, in detail, by dictation, or as free text, and saved sessions appear as regular entries in the progress notes.

No. The six AI actions — case formulation, goal suggestions, treatment plan, homework, session summary and relapse prevention — return editable drafts tailored to the chosen therapy approach. Only de-identified plan context is sent, and nothing enters the record until you review and explicitly accept it.

Yes. The "Konsil" case area composes the request fully locally — specialty, clinical question, short history and selected case sections, de-identified or pseudonymised as you choose. It is handed over by print (cover sheet) or file export with the full content. Online access for external physicians (login/invite link) is deliberately still disabled.

Paste lab report text and the parser extracts values into structured entries, or import CSV/Excel files through the Integration Hub. Values are displayed with reference ranges, trend graphs over time, and optional correlation with medication changes.

Yes. The template library ships with clinical document templates and assessment instruments, and a community library lets you download templates shared by colleagues (with ratings) or publish your own. Templates never contain patient data.

"Patient education — free choice of topic" drafts patient-friendly education material on medications, conditions, therapies or free topics — AI-generated section by section and final only after your approval. While you type the subject, your own existing documents and matching templates from the library and community are offered so nothing gets generated twice. Every output — print, PDF, Word, text, clipboard and the saved note — ends with a consent section (checkboxes plus signature lines for patient and clinician). Approved documents can be saved to the template library, and each document carries created/modified metadata with a revision history. Within the Medication tab, education can also be created directly for the whole plan, for selected medications or for a single drug; if an active drug already has an approved education document in the knowledge base, it is offered there automatically.

Yes — the Integration Hub imports files without requiring technical knowledge: choose a file and the format (Psychiatry.Ink JSON, FHIR bundle, CSV/Excel tables, CDA/XML documents) is detected automatically. You always see a preview with duplicate detection first, and nothing is written to a case without your explicit confirmation.

At any time. Case exports are available as Psychiatry.Ink JSON (complete, re-importable), FHIR R4 bundle, CSV/Excel tables, or a readable PDF package — with a preview of exactly what each export contains. Your data is never locked in.

Yes. Appointments can be exported as a standard calendar file (.ics) that works with Apple, Google, and Outlook calendars. By default entries are privacy-neutral ("Psychiatry.ink appointment") — no patient names or clinical content leave the workspace unless you explicitly choose otherwise.

A guided wizard helps you set up data exchange with hospital, practice, and lab systems, oriented to the standards common in your region. File-based exchange (FHIR, HL7, CSV) works today; direct live connections are prepared but deliberately disabled until secure credential management is in place.

You can export every case completely (JSON, FHIR, PDF) before or after cancelling. Account deletion removes your data from our systems; because identifiers are encrypted with your keys, they were never readable by us in the first place.

Practice mode extends the same application with a team workspace: you invite members via email link, assign roles (admin/psychiatrist, psychologist/psychotherapist, reception/assistant, therapist, viewer), share cases selectively, communicate via team notes, handovers and tasks, and review an activity/audit log. Practice Starter includes up to 4 seats and 1,500 shared AI credits per month; larger plans extend seats, credits and permission depth.

The practice subscription runs on the practice owner's account, whose balance becomes the shared pool: every AI action by an active member is debited from it. The admin can optionally set a monthly credit limit per member, sees per-member usage in the practice dashboard, and gets warnings before the pool runs low. Additional credit packs and auto-recharge work exactly like in single-user mode.

By default only appointments, tasks and administrative data. Clinical notes, diagnoses, medication, lab values, risk assessments and psychotherapy content are excluded — not just in the UI but cryptographically: without a granted case key, clinical content and team notes cannot be decrypted at all. The practice admin can explicitly enable individual permissions (e.g. "view clinical content") per person.

Each shared case has its own AES-256 case key, wrapped individually for every team member with their personal RSA key. The server stores only ciphertext and wrapped keys — it can never read content. When the admin revokes case access, the member's wrapped key is deleted.

The single-user plan starts with a one-month free trial including 500 AI credits, then £24.99/month or £239.90/year (20% saving); Pro Plus with 1,500 credits per month is £49.99/month. For shared practices there are the Practice plans with a team workspace, roles and a shared AI credit pool: Practice Starter (up to 4 seats, 1,500 shared credits/month) at £79.99/month, Practice Team (up to 8 seats, 4,000 credits/month) at £149.99/month, and Practice Plus (up to 15 seats, 10,000 credits/month) at £299.99/month. Enterprise deployments for organisations are in development on request. Details are in the pricing section above.

Nothing is lost and nothing stops working except AI actions: you can keep writing, editing, and exporting without limits. Buy additional credits or wait for your monthly allowance to renew, and AI assistance resumes.

First use the password-reset link on the sign-in page. Make sure you are on the correct workspace address (app.psychiatry.ink) and that your browser allows cookies and local storage for it. If you still cannot get in, reach us through the contact form and we will help promptly.

Patient identifiers are encrypted with keys that live on the device where they were created — that is the zero-knowledge design. To work across devices, enable the encrypted account backup with a passphrase (Settings → Workspace vault) and restore it on the new device; cloud sync tiers keep clinical content available across devices. If individual items — such as medication-education documents or notes — are missing on the second device, click "Sync now" in the top-right of the dashboard; it reconciles case files, identifiers, education documents and notes with the account in one pass.

Inside the workspace, the help assistant (the ? icon in the top bar) answers questions about features and workflows. For anything it cannot resolve — bugs, billing, privacy requests — use the contact form and our team will respond by email.