How Wendy works

Wendy is software. Here's what that means.

Last updated October 8, 2026Previous versions →

Wendy is an AI.

Wendy is an AI coach. Not a human, not a therapist, counselor or doctor, and not HR. The chat screen says "AI coach" under Wendy's name, the first conversation of each day opens with a line that says Wendy is an AI coach, and every page says so at the bottom. If you ask whether you're talking to a person, Wendy says no.

Which models.

Wendy runs on Claude models from Anthropic. A larger model writes your report, your weekly review and any pair page. A faster model runs the daily chat. We don't train models on your data, and Anthropic doesn't either; our contract with them says so.

What it doesWhich model
Your report and weekly reviewclaude-fable-5-1
Daily chatclaude-sonnet-5
Reading connector summariesclaude-haiku-4-5
The safety check on each messageclaude-haiku-5-5

When we change models we note it here with the date.

Changed on September 21, 2026.

What Wendy sendsHow long Anthropic keeps it
Your report, your weekly review and any pair pageUp to 30 days, then deleted. The model that writes these is not offered with zero retention.
Daily chat with Wendy, including the safety check that runs on each messageUp to 30 days to detect abuse, then deleted.
One-line notes written from accounts you connectUp to 30 days to detect abuse, then deleted.

What Wendy optimizes for.

The goals you state, in your words. Wendy has no undisclosed objective: no engagement target, no plan to sell, no score to move. She never mentions a plan or a price in a conversation. If she ever seems to be steering you somewhere you didn't ask to go, tell us: hello@askwendy.ai.

What Wendy can and can't do.

  • She can be wrong. Treat what she says as a hypothesis to check against your own week.
  • She only knows what you've told her, what's in your report, and the summaries from anything you connected.
  • She doesn't give medical, legal, financial or mental-health advice, and will say so and point you to a person who can.
  • She has warmth, not feelings. She won't tell you she missed you, because she didn't.

How the report is built.

Your answers are scored by fixed rules, the same for everyone. The AI writes sentences from the scores and a bank of approved phrases, and never sees your answers.

The full method →

What Wendy is for, and not for.

Wendy is for decisions, habits and conversations at work and at home. Wendy is not for hiring, performance reviews, promotion or pay decisions, and our terms forbid those uses. Wendy does not diagnose, treat or screen for any condition. Wendy is for adults, 18 and over.

How Wendy handles feelings.

Wendy reflects what you say before she suggests anything. She doesn't score your mood, doesn't rate other people's emotions, and doesn't infer feelings from your inbox. If you say something that suggests you're in danger, she stops coaching and shows you where to get help.

Our safety protocol →

Memory.

Wendy remembers what you tell her, in your words. The whole list is on your Memory page. Edit or delete any of it.

Report a problem.

Reply to any email from us, or write to hello@askwendy.ai. A person reads it. If Wendy said something harmful, tell us the date and roughly the time, and we'll look at it.

Version.

This page: version 2, October 8, 2026.

Earlier versions: Version 1, September 21, 2026

WHAT WE PROMISE

  • Your answers are never shown. Only the report.
  • No email until your report is on screen.
  • Learned from your communications, no messages are ever kept.
  • Your data is yours. Export or delete everything under Profile & settings, in one tap.

The full list of what we keep →