$/plugin install betterprompt@betterprompt

See how you actually use AI.

Analyze AI coding sessions from Claude Code and Cursor directly inside Claude Code. Get local reports on thinking patterns, efficiency, and growth areas without a separate analysis server.

Open SourceSelf-HostedYour Data Stays Yours
See how it works
How It Works

From raw sessions to team insights

01

Your team's AI usage, finally visible

See every engineer's AI sessions, token consumption, and efficiency scores in one dashboard. The growth leaderboard ranks your team by improvement trajectory - not just raw output.

Team Dashboard
Team dashboard with growth leaderboard, session counts, token burn rates, and efficiency scores per developer
02

The blind spots your team doesn't know they have

Automatically surface team-wide anti-patterns like prompt inflation, context bloat, and verbose error pasting. See who’s affected and get actionable recommendations.

Blind Spot Detection
Anti-pattern deep dive showing prompt length inflation, context bloat, and other patterns across team members
03

Sprint updates that write themselves

Know who built what across every project - without status meetings. Session data maps to repositories and shows each contributor's work alongside their session count.

Project Activity
Project activity view showing team members, their sessions, and contributions across projects

Analyzes sessions from

Claude CodeCursor
For Teams

Your team builds with AI every day.
Do you know how well?

Team Velocity
142 sessions · 38.4h · 284k tokens this week
NowNo visibility into how engineers use AI - managers guess who is effective based on output alone
FixSession-level analytics: token burn rate, anti-pattern frequency, and efficiency scores per developer
Growth Tracking
8 members · 5 dimensions · weekly snapshots
NowAI training is a one-time workshop. No way to measure if developers actually improved over time
FixContinuous scoring across Thinking, Communication, Learning, Efficiency, and Outcomes with trend lines
AI Retrospectives
Auto-generated from real session data
NowSprint retros mention AI vaguely - "we should use AI more" - with no data
FixData-driven KPT retrospectives showing what to keep, problems to address, and experiments to try
Privacy

Your data stays yours

On Your Machine

Session data stays local during analysis and does not leave your machine by default

Plugin-First

Claude Code runs the analysis locally with the model session you already have open

Queued, Not Detached

Auto-analysis is queued into the next Claude Code session instead of spawning hidden background jobs

BetterPrompt is open source and local-first. Analysis runs inside Claude Code, and team sync is optional if you want dashboard storage.