Sessions

Open Interpreter stores conversations locally under ~/.openinterpreter/ so you can continue work later.

Resume

Resume the newest session for the current directory:

interpreter resume --last

Open a picker:

interpreter resume

Include sessions from other directories:

interpreter resume --all

Resume a known id:

interpreter resume <SESSION_ID>

Fork

Forking creates a new thread from an older session. The original stays intact.

interpreter fork --last
interpreter fork <SESSION_ID>
interpreter fork --all

Inside the TUI, use /fork or /side.

Exec Sessions

Non-interactive sessions can also be resumed:

interpreter exec resume --last "continue with the implementation"

Use --include-non-interactive with interactive resume when you want exec sessions to appear in the picker.

Compact

Long conversations can be compacted into a shorter summary:

/compact

Automatic compaction can also run when the active model is near its context limit. Set model_auto_compact_token_limit in config if you need an explicit threshold.

History Controls

Disable saved transcript history:

[history]
persistence = "none"

Cap history size:

[history]
max_bytes = 104857600

Daemon

interpreter runs in the foreground by default. To share one runtime across several clients, run the app-server as a background daemon and stop it with:

interpreter app-server daemon stop

See the Daemon page for details.