Choose your entrance
There are two safe ways in, and both lead to the same system. Almost everyone starts with Door 1 — no installation, working tonight in the browser AI you already use. Door 2, the full governed Hermes build, is optional and can wait as long as you like.
Use the browser AI you already have (ChatGPT or Claude). Follow the ecosystem guide's four browser-safe steps: take the SOUL Quiz, paste your SOUL file into a no-patient-data conversation, save the Starter Kit on your device if useful, and run one reviewed workflow. The Hermes-only local-folder and dashboard-editing steps are not part of Door 1.
Choose Door 1 when: you want a win tonight, your computer can't run Hermes yet, or you'd rather prove the workflows before committing to new software.
Guide me step by step → Take the SOUL Quiz Use the four browser-safe steps
Hermes is the free, open-source AI engine by Nous Research. It can keep Nurse AI OS files, standing instructions, and local state on your machine. Unless you deliberately use a local model, prompts and selected context may still be processed by your configured model provider under that provider's terms. Install Hermes only from the official site, with the Pre-Procedure Checklist beside you.
Choose Door 2 when: your Door-1 workflows are proven and you want persistent local files and governed, provider-aware workflows on a personal or explicitly authorized machine.
Guide my macOS setup → Start Here, Step 1 Hermes Install Guide Pre-Procedure Checklist
The rules that protect everyone
You memorized the rights of medication administration. These are shorter, and they protect your patients, your license, and you — behind either door, in any AI tool.
No names, charts, screenshots, room numbers, record numbers, dates, or stories that could identify a patient — in any AI tool, at any level. If you're unsure whether something counts: it counts. Leave it out.
Every AI output is a draft until a human reviews it. Nothing gets sent, published, or acted on without your judgment.
No AI here supports clinical decisions, assessments, or patient-specific care questions. It carries your busywork — never your judgment.
No EHR screens, no confidential policies, no identifiable staff or student data. Public, fictional, or personal material only — unless your organization's compliance team has explicitly approved a tool and workflow.
Bring gray areas to the Lamp Huddle or email robert@nurse-ai-os.org. A question asked early beats an incident reported late.
These rules have a name: EDENA — the governance, risk, safety, and assurance control plane — the governance framework in every layer of Nurse AI OS. In the current public setup, EDENA provides review-first classifications for human judgment: 🟢 Green — low-risk and reversible 🟡 Yellow — human review before action 🔴 Red — stop, prohibit, or escalate The Governance Kit provides advisory instruction files; it does not by itself enforce controls. The published default posture remains shadow/observe-only, with humans responsible for every action.
Use the Setup Helper—with local-only progress—for one verified step at a time, or read the complete five-step walkthrough on one page.