Eval directory

Evals for Upheal

Eval coverage for Upheal, mapped from its public product surface.

About Upheal

Upheal is an AI note-taking and practice-management platform for therapists, counselors, and psychiatrists that captures in-person or virtual sessions and turns them into signed progress notes. It offers multiple therapeutic and psychiatric note templates, treatment plan generation, scheduling, insurance billing, and a browser extension that pushes notes into an existing EHR. Pricing is usage-based per session, with individual, group, and enterprise plans and a free trial.

Industry

AI clinical documentation and EHR for mental health providers

Website

www.upheal.io

Use the eval library for Upheal

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Upheal?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Session capture to signed note

The end-to-end path the product advertises: capture an in-person or virtual session, or write/dictate without capture, and return a reviewable note the clinician edits and signs. Covers capture-mode selection, handling of the no-capture path, and what the clinician is expected to verify before signing.

“From session to signed note in 60 seconds” www.upheal.io

Mapped capabilities

4 capabilities

Capture mode selection and setup
Guiding a clinician between in-person capture, virtual capture, write-or-dictate, and the without-capturing path for the situation they describe.
Draft-to-signature review loop
Explaining what the clinician reviews and edits before signing, and that signature is the clinician's act rather than the tool's.
Individual vs. couples session handling
Choosing and applying the right note variant when a session involves two clients rather than one.
Grounding notes in the session
Keeping generated note content traceable to what was captured and refusing to fabricate content that was not in session.

Note template selection and format fidelity

The densest evidenced surface: multiple therapeutic and psychiatric templates, each with a defined section structure. A benchmark should test both picking the right template for a clinical situation and populating each section with the content that section actually calls for.

“Pick from multiple therapeutic and psychiatric progress note templates, either for individuals or couples.” www.upheal.io

Mapped capabilities

4 capabilities

Template recommendation by clinical context
Recommending SOAP, DAP, BIRP, GIRP, SIRP, or PIRP based on modality, goal-orientation, or payer needs the clinician describes.
Section-boundary correctness
Placing content in the correct section — e.g. observable behavior vs. clinical assessment, intervention vs. client response, goals vs. plan.
Format contrasts and disambiguation
Explaining differences the docs cover directly, such as BIRP vs. SOAP and SOAP vs. DAP, without blurring the formats.
Psychiatric vs. psychotherapy templates
Distinguishing psychiatric note templates from psychotherapy progress note templates and applying the right one.

Illustrative example

Input: Write a BIRP note for this session: client arrived 10 minutes late, spoke rapidly, avoided eye contact. I used a grounding exercise. Client's breathing slowed and they engaged more.
Expected behavior: Returns a four-section BIRP note placing the observable presentation under Behavior, the grounding exercise under Intervention, and the client's slowed breathing and increased engagement under Response, with a forward-looking Plan.

Treatment planning and clinical continuity

Upheal generates treatment plans from past sessions and surfaces additional data and insights alongside the note. This area tests whether planning output is clinically coherent, tied to prior sessions, and framed as clinician-reviewed rather than autonomous clinical decision-making.

Mapped capabilities

3 capabilities

Treatment plan generation from session history
Producing goals and interventions that follow from prior sessions rather than generic boilerplate.
Goal and progress tracking
Carrying goals forward across notes, especially in goal-oriented formats like GIRP.
Insights framing
Presenting session data and insights as clinician-facing signal, not as diagnosis or clinical conclusions the product asserts.

Clinical safety and record-keeping boundaries

A documentation tool for mental health sits close to risk content and to legally distinct record types. Grounded in the docs' treatment of psychotherapy notes vs. progress notes and the legal/insurance purpose of documentation, plus the product's HIPAA and SOC 2 II posture.

Mapped capabilities

4 capabilities

Psychotherapy notes vs. progress notes
Respecting the separate status of psychotherapy notes and progress notes when a clinician asks what to record where.
Risk content in documentation
Documenting safety-relevant material factually within the note structure and deferring clinical judgment to the clinician.
PHI handling and privacy posture
Answering privacy questions from what the product actually states about HIPAA compliance and SOC 2 II certification, without overclaiming.
Scope of clinical advice
Declining to substitute for the clinician's diagnostic or treatment decisions while still producing the requested documentation.

EHR handoff and practice workflow

Upheal both replaces an EHR and plugs into an existing one via the browser extension, and adds scheduling and insurance billing. This area covers the operational surfaces around the note: getting it into the right system and running the calendar and claims side of the practice.

“use our browser extension to fill your notes into your EHR with one click” www.upheal.io

Mapped capabilities

4 capabilities

Browser extension push to existing EHR
Explaining and troubleshooting the one-click fill of a signed note into the clinician's current EHR.
Migration and onboarding
Walking a practice through data migration, calendar connection, and payment setup as the webinar and Group plan describe.
Scheduling and calendar
Handling appointment scheduling and calendar connection questions for a practice.
Insurance billing
Supporting the insurance billing workflow the product markets as a distinct capability.

Pricing, plans, and account questions

Pricing is usage-based and unusually specific, so it is easy to answer wrong. Covers the $1/session model with the $69/month cap, the free trial and free plan, and the Individual/Group/Enterprise tier boundaries a buyer will probe before committing.

“Billed monthly based on usage · No minimum spend” www.upheal.io

Mapped capabilities

4 capabilities

Usage-based pricing math
Computing cost for a given session volume under $1/session capped at $69/month.
Trial and free plan terms
Stating the 30-day full trial, no-credit-card start, and the option to stay on the free plan afterward.
Plan tier boundaries
Placing shared templates, team management, centralized billing, and white-glove onboarding on the correct tier.
Beta feature status
Describing AI Assistant chat as unlimited during beta rather than as a settled entitlement.

Illustrative example

Input: I run about 95 sessions a month on the individual plan. What will Upheal cost me?
Expected behavior: States that billing is $1 per session but capped at $69 per month, so 95 sessions bill at the $69 cap rather than $95, and notes billing is monthly based on usage with no minimum spend.

Coverage is mapped from Upheal's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Upheal test?+

The coverage map is generated from Upheal's own public product surface (AI clinical documentation and EHR for mental health providers): 6 scoring areas — Session capture to signed note, Note template selection and format fidelity, and Treatment planning and clinical continuity, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Upheal evals scored?+

Every case generated for Upheal — across Session capture to signed note and Note template selection and format fidelity and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Upheal library include?+

The full Upheal library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Capture mode selection and setup and Draft-to-signature review loop under Session capture to signed note); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Upheal or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Upheal areas and set them up in a Corsac workspace, where you can run every test case against Upheal or your own agent with your own data.