Skip to content
nomadv0.62.6

Privacy

claude-nomad runs entirely on machines you control and the private Git repository you configure. The project operates no server, no backend, and no analytics, and the author receives no data about you, your machine, or how you use the tool.

  • No telemetry, analytics, tracking, or usage metrics.
  • No account, login, or identifier of any kind.
  • The author and the project receive nothing about you or your Claude Code configuration.

Your Claude Code configuration and session transcripts sync only between your own machines and the private Git remote you configure (your NOMAD_REPO / GitHub repository). That remote is infrastructure you own and control. claude-nomad never copies your configuration anywhere else.

If nomad hits an unexpected bug, it writes a crash report to ~/.cache/claude-nomad/crash/ on the local machine only. The saved report is owner-readable-only (0o600) and redacted: a structural pass replaces your home directory and hostname, and a best-effort gitleaks scan then strips detectable secrets. That secret scan runs against a temporary, owner-only (0o600) copy that already has the structural pass applied; the scratch file is deleted right after the scan, so only the fully-redacted report remains. Nothing is ever transmitted or auto-uploaded; the report exists solely for you to optionally attach to a bug report.

nomad doctor makes at most two read-only outbound requests. Each is a plain GET that fetches public data and is bounded by a three-second ceiling. Neither sends any information about you or your configuration:

  • npm registry (registry.npmjs.org): on every run, reads the latest published claude-nomad version so the doctor can warn you when a newer release is available. The version comparison happens locally.
  • JSON Schema Store (json.schemastore.org): only when you pass --check-schema, fetches the public Claude Code settings schema used to validate your local settings.json. The default nomad doctor, and the plugin’s session-start hook, never makes this request.

The doctor also prints your configured Git remote, but that is a local read of .git/config (via git remote get-url origin), not a network request. Sync to that remote happens through git only when you explicitly run nomad pull or nomad push, against infrastructure you own and control.

Like any HTTP request, the two requests above reveal standard connection metadata (such as your IP address) to the endpoint you contact (npm or the schema store). That is inherent to making a network request; it is not collected by claude-nomad or its author. You can run the tool fully offline, in which case these checks simply report as unavailable.

The companion plugin adds slash commands and a session-start hook that shell out to the local nomad CLI. The hook runs the default nomad doctor, so its only outbound network request is the npm version check described above. The plugin introduces no data collection or network activity of its own.

This policy may be revised as the tool changes. The authoritative version is the one published with the current release on this site.