YOUR WORDS. YOUR CALL.

Good feedback.
A little personality.

Set your preferences, try a spicy prompt, or give Jev a turn. You’re always in charge of Send.

Make yourself comfortable.

Local checks run only on the sites you enable. Reload open chats after installing or updating.

FEEDBACK PERSONALITY

OPTIONAL · JEV SECOND OPINION

Meet the other tone detective.

Jev adds contextual classification. It does not write a new message. Your draft goes to your configured relay and TypeSafe AI only after you confirm a check.

Never paste your Jev API key here. It belongs in your relay’s server environment. Enter the separate relay access token below.

Local mode · no cloud connection

Allow this origin on your relay:

How to start the relay

See Jev setup instructions. The key stays server-side; requests are authenticated, bounded and rate-limited. This beta’s relay is intended for local use or a small private test.

Fixed test sample (only this text is sent):

That answer is incorrect. Check the calculation.

THE PLAYGROUND · NOTHING IS SENT TO A CHAT

Go on. Give it some attitude.

Click the meter to see the explanation and preview a targeted edit. Copy it, review it, then paste it yourself.

PRIVACY, WITHOUT THE FINE-PRINT ENERGY

Your draft is your business.

  • Local checks use only the current composer. No chat history, cookies or account information.
  • No telemetry, advertising or saved drafts. Preferences stay on this device.
  • Jev is optional and never runs automatically. Each check shows the draft and destination first.
  • Relay tokens live in extension session storage, inaccessible to website content scripts. Reconnect after restarting Chrome.
  • Your relay and TypeSafe have their own data-handling policies. This extension cannot promise what an independently hosted relay stores.
Full privacy details ↗

You’re allowed to disagree.

“That is incorrect. Check your sources.” is a useful instruction. You don’t owe your autocomplete a compliment sandwich.

Local checks are English-first heuristics. They can miss sarcasm, dialect and context. Neither green nor a Jev verdict predicts policy compliance or account safety.

What about the Claude policy?

Anthropic announced an update effective November 12, 2026 addressing sustained, needless abuse. It excludes ordinary frustration and pushback, and identifies ending abusive conversations as the primary enforcement mechanism. An ended chat is different from an account ban.

Read Anthropic’s explanation ↗ · Checked October 11, 2026.