Skip to content

Data & Privacy ​

Data is 100% local. Nothing ever leaves your machine.

AssetPlex is a local-first tool with no cloud account, no server, and no telemetry endpoint. The full policy lives in the repository at PRIVACY.md.

Where your data lives ​

All AssetPlex data is stored in a single directory on your own machine:

~/.assetplex/

This includes your identity profile, skills, rules, MCP server configurations and preferences.

What AssetPlex reads ​

When you run assetplex doctor, assetplex sync or the import wizard, AssetPlex reads your AI tools' local configuration files (e.g. ~/.claude/, ~/.codex/, ~/.trae-cn/) — only to detect tools, sync your hub content, or import existing configs. It never scans your whole disk or collects usage statistics.

What AssetPlex writes ​

AssetPlex only writes to:

  1. Its own hub directory ~/.assetplex/
  2. The tool config files you explicitly choose to sync or import (as configured in hub.toml)

Every sync action is opt-in per tool. Use assetplex sync --dry-run to preview changes first.

MCP configurations may contain secrets ​

MCP server configs can contain API keys or tokens. AssetPlex processes these locally only — it never transmits them, never logs their contents (logs record only file paths and operation results), and you fully control which configs sync to which tools.

What AssetPlex never does ​

  • No analytics, telemetry or usage tracking
  • No account or login required
  • No cloud sync or backup of your data (backups, if enabled, are local in ~/.assetplex/.backups/)
  • No sharing of your data with any third party

Released under the MIT License.