Support

Drop a note below. Typical response same business day. If something is broken mid-chart, send and keep working — Cortex can be paused or removed at any time without touching the EMR.

Cortex Lens is currently distributed two ways:

• Chrome Web Store (unlisted): if we sent you a store URL, click Add to Chrome and follow the prompts. Updates land automatically.
• Direct zip: if we sent you a .zip file, follow the steps in the install one-pager we sent alongside it (also available in the repo as extension/INSTALL.md). Updates require swapping the unzipped folder and reloading the extension on chrome://extensions.

Cortex Lens currently supports Practice Fusion (chart pages on static.practicefusion.com). More EMR adapters land as the pilot expands.

Most issues fall into one of these patterns:

• Card doesn't appear on a chart. Click the extension toolbar icon. Look at Adapter health:
  • no chart on page → you're on a non-chart view (login, schedule, billing). Open a patient.
  • broken → the EMR's page structure changed and Cortex hasn't caught up. Email us; we have a next-business-day fix path for these.
• "Cortex Lens was updated — reload this tab" banner inside the card. Reload the EMR tab; that fixes it. (This appears after Chrome installs an update while a chart tab is still open.)
• Analyze returns an error. Click Debug → Copy scan JSON in the popup, paste into your email to us. The diagnostic includes everything we need to reproduce.
• Pause Cortex during a sensitive encounter. Open the popup → Snooze. Choose a duration (5 min through until-tomorrow). The card disappears entirely until the timer expires or you re-enable.
• Permanent off. Popup → Off toggle. Or remove the extension from chrome://extensions.

The extension reads structured chart context (problem list, medications, recent labs, demographics) and the in-progress note text from the EMR page you're viewing. It strips patient name and PRN/MRN before any network call. The structured context is sent to Cortex's analysis API, which forwards to a frontier large-language model under our review.

The pilot operates on demo or de-identified data only while the formal posture (BAA, zero-data-retention with the LLM provider, signed pilot agreement) is in progress. Don't use Cortex on a chart you wouldn't be comfortable putting on a research server until that work lands.

Detailed methodology and benchmarks are public: cortexcharts.com/benchmarks. Privacy posture: cortexcharts.com/privacy.

Pilot feedback is the highest-priority input we have right now. A note that says "this would have caught X but missed Y" — even one line — beats any internal benchmark. Email anything, any time.