Getting started

Install

Download the command line (and the MCP server) or the desktop application for your platform, unpack the archive and put agentlens on your PATH.

First commands

agentlens scan                       # which sources exist, how many sessions
agentlens skills --since 30d         # skill usage and token cost
agentlens unused --days 30           # installed skills, agents and MCP servers nobody used
agentlens cost --since 30d           # cost per model, measured vs. computed
agentlens report --html report.html  # one self-contained page that works offline

scan comes first: it tells you which of the four sources AgentLens found and how many sessions it read. Everything else works on those sessions.

Narrowing down

Every command takes the same filters:

Flag	Meaning
--since <WHEN>	Only sessions active since a point in time: a duration (12h, 30d, 4w) or a date (2026-10-01).
--source <ID>	Only this source (repeatable): claude-code, opencode, copilot-cli, vscode-copilot.
--project <DIR>	Only sessions that ran in this project directory or below it.
--format <FORMAT>	table (default), json or csv; see Reports and export.
--otel <FILE>	Add an OpenTelemetry export to the analysis (repeatable).
--rounded	Token counts in text tables as 1.55 M, 2.11 k or 3.20 B instead of the full number; JSON and CSV stay exact.

Reading the numbers

A leading ~ marks an estimate; a value without it was reported by the client; n/a means the client does not record it. Measured and estimated explains the rule, and the analyses explain the columns.