Skip to main content
Tuple keeps a local index of every call you’ve transcribed, so you can easily reference what you talked about on past calls. The index lives in a SQLite database that the Tuple daemon writes as each call records, and you query it from the tuple CLI.Everything stays on your machine - the transcript database is local.

​
When to use this

Reach for tuple transcription when you want to:
  • Find the call where you and your pair discussed a specific decision, error message, or library name.
  • Pull up the transcript of a recent call without remembering exactly when it happened.
  • Export transcripts to Markdown so you can drop them into your team’s notes or feed them to an LLM.

​
Listing recorded calls

Use tuple transcription list to see calls in the database, most recent first:
Each row shows the call’s start time (in local time), a short call ID, the number of transcript segments captured, and the participants Tuple resolved from the recording.Filter the list with any combination of:Example — every call you had with someone named Mikey in the last month:
Pass --format json (a CLI-wide flag) to get the same data as JSON for scripting.

​
Searching transcripts

tuple transcription search runs a full-text search across every indexed transcript:
The query uses SQLite FTS5 syntax: bare terms are ANDed together, "quoted phrases" match exactly, and OR and NOT work as you’d expect. Matched terms in the output are highlighted with [[ ]] markers turned into bold text.Each result shows the call ID, timestamp, speaker, and a snippet around the match. Refine results with:Example — find every time your teammate mentioned “rollback” with two lines of context on either side:

​
Showing a full transcript

Once you’ve found the call you want, pass its short ID to tuple transcription show to print the full transcript:
The ID can be any unique prefix of the full call ID — the same short IDs list and search print. The output is a sequence of timestamped speaker lines. Add --format json for the raw segments.

​
Exporting transcripts

tuple transcription export writes one file per call into a directory, named <date>@<call-id>.<ext>:
Choose the output format with --format:Filter which calls get exported with the same --call, --participant, --after, and --before flags as list. Calls without transcript segments are skipped, and existing files are overwritten — re-running the command refreshes the export.Example — export every transcribed call from May 2026 as Markdown: