Therapeutical AI

Avoid burnout. Waste fewer AI tokens.

Therapeutical AI predicts the risk of unproductive AI work or increasing mental strain over your next two hours.

It combines how you feel with token use, repeated attempts, and review activity to give you one experimental risk percentage and a clear next step:

Continue · Clarify your goal · Take a break · Stop for today

AI that learns from your work

Four-question check-ins appear about every two hours while you work. They connect your current state with what happens afterward, building a personal history for predictions.

Your first seven active days help establish your patterns. Early suggestions use simple rules. The AI model starts making predictions once enough completed follow-ups are available.

Everything you need to act

• Token tracking: see usage and reported costs from connected AI tools.
• Unread AI changes: find generated code that needs review.
• Shareable summary: see your session or day in one place, then save or copy an image.

Get started

Install, complete setup, and open Therapeutical.ai: AI work summary.

To connect GitHub Copilot, run Therapeutical.ai: Connect AI tool, select GitHub Copilot (VS Code / CLI), and start a new agent session after connecting. Use a current VS Code or Copilot CLI release with agent hooks enabled.

If VS Code does not pick up the connection, check that chat.useHooks is enabled and ~/.copilot/hooks is enabled in chat.hookFilesLocations.

The Copilot connector tracks local agent turn boundaries and context compactions and works with the unread AI code queue. It does not track inline completions, cloud-agent runs, or Copilot token/cost totals.

Free research preview · No user API key required

Local tracking works without cloud participation. Cloud predictions require consent to share numeric activity and check-in data. Code, prompts, and file names are never uploaded.

Predictions are experimental, not medical assessments.

Website · Privacy policy