Actualis

Documentation

Documentation

Actualis reads the logs your coding agents already write and turns them into answers about exposure, activity and cost. It runs on your machine, opens your transcripts read-only, and makes no network connection.

These pages cover the free tool completely. There is nothing held back for a paid tier — every detector, rule and command documented here is in the version you can install right now.

Install

One command. Python 3.9+, no dependencies.

Your first report

Run it, and understand what came back.

Command reference

Every flag, what it does, when to reach for it.

What it detects

Credential types, shell audit categories, coach findings.

Before you start

Actualis reads transcripts that already exist. If you have never run Claude Code or Codex on this machine, there is nothing for it to read and it will tell you so rather than pretending otherwise.

AgentWhere its transcripts live
Claude Code~/.claude/projects/**/*.jsonl, plus $CLAUDE_CONFIG_DIR
Codex$CODEX_HOME/sessions/**/rollout-*.jsonl

Those are the only paths it opens. You can confirm that yourself — see verifying the privacy claims, which is a page of commands rather than a page of promises.