Skip to main content

​
Install as an agent skill

If you work inside Claude Code, Codex, Cursor, or another Agent Skills host, follow Install a reviewed revision. That procedure fixes the installer version and requires an operator-supplied, reviewed full public-repository commit SHA. Agent and installation-scope choices remain interactive. Without a reviewed SHA, stop the installer step or import an already-reviewed local folder after permission; current-session use remains separately permission-gated. The agent runs the CLI itself and explains the findings. See skill bundle and agent clients for details.

​
Supported Node runtime

@xerg/cli supports Node 22.13.0 and newer. That is the first Node 22 release where node:sqlite is available without its startup flag.

​
Supported data sources

  • local OpenClaw logs and session transcripts, plus optional independent sanitized trace captures
  • local Hermes state.db, with legacy log/transcript fallback where present
  • local Claude Code session transcripts
  • local Cursor usage CSV exports
  • event payloads from any framework via xerg ingest
  • OpenClaw SSH, Railway, and multi-source remote audits

​
Run without a global install

Direct commands remain available when you want explicit control:

​
Install globally

​
Runtime note

@xerg/cli is a Node.js CLI. Use Node 22.13.0 or newer, then run npx @xerg/cli@latest init to confirm the package is working.

​
First-run defaults

xerg init is the default first-run path. It is interactive and best when you want the fastest local result. Use direct doctor / audit commands when:
  • you need non-interactive behavior
  • you want JSON or Markdown output immediately
  • you want to script specific runtime or path flags
  • you want remote OpenClaw audits directly

​
Develop from this repo

If you are working from the Xerg repo itself:
This repo pins the default development and release toolchain to Node 24.14.0 in .nvmrc.