CLI & Agent Integration

Local Transcriber ships with a CLI tool and an agent skill for AI coding agents like Claude Code and pi. Install both with a single menu item.

Installation

In the app menu bar, go to Transcriber → Install CLI & Agent Skill…

This does two things:

1. CLI — symlinks transcriber to /usr/local/bin/ (requires admin password).
2. Agent skill — copies the skill file to ~/.pi/agent/skills/transcriber/ and ~/.claude/skills/transcriber/.

After installing, open a new terminal and run:

transcriber help

CLI Commands

transcriber start

Start a new recording. Launches the app if it’s not running.

transcriber start
# ✅ Recording started
# {"recording": true, "status": "Transcribing…", ...}

transcriber stop

Stop the current recording and begin post-processing.

transcriber stop
# ⏹  Recording stopped
# {"recording": false, "status": "Processing…", ...}

transcriber toggle

Toggle recording on or off.

transcriber toggle

transcriber status

Print the current status as JSON. Works whether or not a recording is active.

transcriber status
# {
#   "recording": true,
#   "status": "Transcribing…",
#   "duration": "05:32",
#   "segments": 24,
#   "file": "/Users/you/Documents/MeetingTranscripts/2026-04-03-1400.md",
#   "partial": "and then we should probably…",
#   "updated_at": "2026-04-03T14:05:32Z"
# }

Fields:

Field	Description
recording	Whether a recording is active
status	Current state: Ready, Transcribing…, Processing…, Loading models…
duration	Session length as MM:SS
segments	Number of committed transcript segments
file	Path to the current .md transcript file
partial	Uncommitted partial text (live, not yet finalized)
updated_at	ISO 8601 timestamp of last status update

transcriber transcript

Print the current or most recent transcript file to stdout.

transcriber transcript

transcriber tail

Tail the current transcript file — output updates live as new segments are committed.

transcriber tail
# Tailing: /Users/you/Documents/MeetingTranscripts/2026-04-03-1400.md
# (new lines appear as the meeting progresses)

transcriber open

Open the Local Transcriber app window.

transcriber open

URL Scheme

You can also control the app via URL scheme, useful for shortcuts, automation tools, or scripts:

open transcriber://start    # Start recording
open transcriber://stop     # Stop recording
open transcriber://toggle   # Toggle recording on/off
open transcriber://pause    # Pause active recording
open transcriber://resume   # Resume paused recording

Note: pause and resume are only available via URL scheme, not the CLI.

Reprocess a recording

Re-run post-processing (improved transcription + speaker diarization) on an existing recording:

open "transcriber://reprocess?file=2026-04-03-1400"

The file parameter is the filename stem (without extension) from your transcript folder.

Agent Skill

After installing, AI agents that support skills (Claude Code, pi) automatically know how to control Local Transcriber. The skill teaches them to:

• Start and stop recordings
• Check recording status
• Read the live or completed transcript
• Reprocess existing recordings
• Find and search transcript files

Your agent reads the transcript directly from disk — no API keys, no cloud, no copy-pasting.

Example: Claude Code

> start recording my meeting

✅ Recording started.

> what's been discussed so far?

Based on the live transcript, the team discussed…

> stop recording and summarize

⏹ Recording stopped. Here's a summary…

Status File

The app writes its current state to ~/.transcriber/status.json on every change. The CLI reads this file, and your own scripts can too:

cat ~/.transcriber/status.json | python3 -c "
import json, sys
d = json.load(sys.stdin)
print('Recording' if d['recording'] else 'Not recording')
"

Engine Log

For debugging, tail the engine log:

tail -f ~/.transcriber/engine.log