Skip to content

Using NativePi

Sessions and storage

Pi session files are the conversation source of truth. NativePi keeps only the desktop state that Pi does not own.

Pi data

~/.pi/agent
├── sessions/      # conversations shared with the Pi CLI
├── packages/      # installed Pi packages
├── settings.json  # agent configuration
└── auth.json      # provider credentials managed by Pi

NativePi opens and updates conversations through Pi's session formats and APIs. You can resume the same session from the Pi command line, and a session created in the command line appears in NativePi's project history.

Session actions

NativePi can create, resume, rename, clone, fork, delete, import, and export sessions. It also presents Pi's session tree, statistics, and compaction workflows. Forks and clones remain ordinary Pi sessions, not NativePi-specific copies.

NativePi data

NativePi persists only interface state:

  • Pinned projects and chats
  • The last open project and chat
  • Unsent text drafts
  • Favorite models
  • Pane sizes
  • Appearance, notification, and keyboard-shortcut preferences

It does not put conversations or credentials in that state file and does not send NativePi-owned desktop telemetry. Pi's optional analytics remain a Pi setting.

External changes

Because NativePi and the CLI share files, do not actively write to the same session from two processes. NativePi detects external changes where continuing could overwrite newer data and fails conservatively instead of silently replacing them.

Related guides