AgentLens offline

Habits you repeat — ranked by what they cost you

Daily spend

Cost by model

By agent

Tool mix

Sessions by cost

Click a session to open its turn-by-turn drill-down.

Searches & greps

Loop-family cost

What's actually in that number

The loop-cost tile above is a total. These are the specific findings it's made of — the same prompts, errors and per-attempt breakdowns as the Waste tab, filtered to just the loop-family detectors.

Prompt size over time

Prompt quality over time

Heaviest prompts

Rules from your own history

Generated from your findings, ranked by what they cost you. Paste straight into your project.

General reference

Static best practices, not derived from your data — worth keeping regardless of what the rules above find.

Suggested for this project

Every suggestion names the finding that triggered it. Nothing here is generic advice.

Generic best practices

Not derived from your data — useful for most projects regardless of what we found above.

AgentLens never writes to your filesystem from the browser. Preview and download here, or run the shown command in your terminal.

Prompt

Scored locally as you type. Nothing is sent anywhere.

All analysis runs on this machine. Nothing is uploaded.

Loading your session history…