Honest, specific, and within clear limits.
Northstar uses AI to read your lab results and vitals, explain each value in plain English against a reference range, surface gentle trends over time, and prepare questions and summaries for your doctor. Here is exactly how, and where the guardrails are.
From your result to your summary.
Each step has a specific job, a specific limit, and a check that keeps it inside that limit.
Reading your data
Understanding labs, vitals, and notes.
Plain-English explanations
Each value explained against a reference range.
Trends over time
Patterns across your earlier results.
Discussion points and visit summary
Grounded in what you submitted. Nothing invented.
Strict guardrails
The things Northstar will never do.
Reading your data
Understanding labs, vitals, and notes.
When you add data to Northstar, a language model reads it to identify what each value is, which unit it is in, and any context you included (such as how you felt, or whether you were fasting). This step is about recognition: turning raw text, pasted lab results, or an uploaded PDF into a structured set of named values.
Northstar does not assume or guess. If a value is ambiguous or the unit is unclear, it asks you to confirm rather than filling in a meaning. Precision at this step matters because the explanation you receive depends on it.
Plain-English explanations
Each value explained against a reference range.
Once your values are identified, Northstar explains each one in plain English: what it measures, what it reflects about your physiology, and where your result sits relative to a general reference range. The explanation is generated by a language model and then validated structurally (using Zod) to confirm it is consistent and does not contain diagnostic claims.
Reference ranges in Northstar are general population ranges drawn from widely published clinical guidelines. They are a starting point for understanding, not a clinical ruling. Your specific lab, your age, your clinical context, and your clinician's judgment determine what any result means for you. Northstar says this clearly next to every explanation.
Trends over time
Patterns across your earlier results.
When you have more than one result for the same marker, Northstar compares them over time. It surfaces the direction of movement (rising, falling, stable) and notes how long the window covers. For example: HbA1c was 5.7% three months ago and is 5.9% now, a modest upward movement over one quarter.
Northstar does not interpret what a trend means clinically. It shows you the pattern and flags it as something worth raising with your doctor. The trend view is designed to give your clinician something concrete to work with, not to lead you toward a conclusion on your own.
Discussion points and visit summary
Grounded in what you submitted. Nothing invented.
Before your visit, Northstar prepares one or two discussion points and a visit summary. Every item in both is traceable to values you actually submitted. If a result is out of range, that is the basis for the discussion point. If a trend is present, that is what gets surfaced. Northstar does not speculate or add context you did not provide.
The visit summary is a structured export: your recent values, any out-of-range flags, a trend view where available, and the prepared questions. It is designed to be shared with a care team member, so it is factual, concise, and free of interpretation. Every AI output in the summary passes through structured validation before it reaches you, ensuring consistency across entries.
Strict guardrails
The things Northstar will never do.
Every AI response Northstar generates passes through a validation layer before it is shown to you. This layer checks that the output does not contain a diagnosis, a treatment recommendation, a medication name as advice, or a dosing instruction. If any of those are present, the response is discarded and you see nothing. This check runs in code, separately from the model itself.
If a value you enter or a symptom you describe matches a pattern associated with urgency (for example, a very high blood pressure result, or chest pain mentioned in a note), Northstar shows you a calm, clear prompt to contact your care team or seek emergency care. It will not attempt to interpret the symptom, suggest a cause, or tell you to wait.
Northstar routes requests across multiple language models and records which model produced each explanation. As better models become available, Northstar can switch without disrupting your data or experience. Your results and explanations are stored separately from the model layer, so your history is preserved regardless of which model is in use. Your data is not used to train public AI models.
Care and safety
A full account of what Northstar will and will not do, and how it handles urgent values and symptoms.
Read the care pageSecurity and privacy
How your health data is stored, encrypted, and kept yours. No third-party sharing without your consent.
Read the security pageCommon questions about the AI.
Try it with your own lab results.
Free to start. Add one value or a full panel. See what Northstar explains.