Chrome Extension

Know when your AI conversation needs a fresh start.

ContextVitals reads the ChatGPT or Claude conversation you're already in and estimates how much context pressure it's under — before the model starts losing track.

The problem

Long conversations quietly degrade.

Early instructions fade

The rules you set at the start stop being followed as the conversation grows.

Important context gets buried

Key details get lost under everything that came after them.

Answers start contradicting earlier ones

The model disagrees with a decision it already made.

There's no clear signal to start over

Message count alone doesn't tell you when a conversation has gone stale.

How it works

One badge, four jobs

Live in v0.2

Context Meter

Live estimates of message count, token usage, and context window pressure, right on the page.

Live in v0.2

Context Health Score

A single 0–100 risk score, weighing token pressure, instruction decay, topic drift, contradiction risk, and complexity together.

Coming soon

New Session Alerts

A quiet nudge once risk crosses into High Risk — not a popup for every message.

Coming soon

Context Handoff

Generates a continuation prompt — goals, decisions, and next steps — so you can carry the important parts into a fresh chat.

The score

Read it at a glance

0–50 Good

Plenty of room left.

51–70 Moderate

Getting long — worth keeping an eye on.

71–85 High Risk

Older instructions and context are under real pressure.

86–100 New Session Recommended

Time to start fresh and carry the important parts over.

Privacy

Built to never see your conversation

Everything runs on your device

Parsing, scoring, and estimates all happen locally in the browser — nothing about your conversation is sent anywhere.

No account, no server

ContextVitals has no backend that your chats pass through.

Every number says "estimated"

ContextVitals can't see a model's real internal context state, so it never claims to — only ever estimates.

Status

What's included

  • ChatGPT and Claude support
  • Message count and token estimates
  • Full Context Health score — token pressure, instruction decay, topic drift, contradiction risk, complexity
  • English, Korean, and Chinese UI

Try it on your next long conversation.

Download for Chrome