The abschrift command line
Every Abschrift install ships a full command-line tool inside the app bundle. It runs the same pipeline on the same vault as the app, with transcription and speaker separation fully on-device. You can process recordings, import files, search your meetings, and automate housekeeping from scripts, cron jobs, or the terminal.
Install
In Settings → Command line, click Install. That links the bundled binary
into /usr/local/bin — macOS may ask for your administrator password, because
that folder is system-owned. The link survives app updates, and Remove in the same
pane undoes it cleanly.
Prefer the terminal? The same thing by hand:
sudo ln -sf "/Applications/Abschrift.app/Contents/MacOS/abschrift" /usr/local/bin/abschrift
Or skip installing and call the bundled binary by its full path — it's the identical program:
/Applications/Abschrift.app/Contents/MacOS/abschrift --help
What it can do
| Command | What it does |
|---|---|
process <mic.wav> <system.wav> | Run the full pipeline on two audio tracks — a new meeting appears in the app. |
process-dir <meeting-dir> | Re-run the pipeline on an existing meeting in place. --force regenerates transcript and notes from kept audio. |
import <files…> | Bring a recording, an exported transcript, or notes from another tool into the vault — same rules as dropping them on the app. --dry-run previews the plan as JSON. |
meetings | List the vault's meetings as JSON, newest first. |
search <query> | Full-text search over notes and transcripts with ranked snippets — the app's sidebar search, for scripts. |
speakers list|rename|delete|merge | Manage speaker profiles. Renames flow back into past transcripts and notes, exactly like in the app. |
models status|download | Check or fetch the on-device speech models (~0.5 GB, one-time). The app's first launch normally already did this. |
llm status|download|list | Manage the built-in local notes engine: hardware tier, recommended model, downloads. |
retention-sweep | Run the retention sweep now (the app also runs it automatically every few hours). |
clean --yes / purge --yes | Bulk maintenance: delete audio+transcripts everywhere, or everything. Irreversible — hence the explicit --yes. |
license status|activate|deactivate | Manage your Abschrift Pro license on this Mac. |
mcp | Start the local MCP server on stdio for AI agents. (Pro) |
abschrift <command> --help shows every flag. Results are JSON on stdout.
Progress and logs go to stderr — pipe-friendly by design.
Examples
Import a downloaded meeting recording and let the built-in local engine write the notes:
abschrift import ~/Downloads/team-sync.mp4 --provider builtin
Find what was decided, from the terminal:
abschrift search "pricing decision" --limit 5
List meetings and pick fields with jq:
abschrift meetings | jq -r '.[] | "\(.started_at) \(.title)"'
To regenerate notes after switching providers, delete the meeting's notes.md
and re-enter the pipeline. Completed stages are skipped:
rm ~/Abschrift/meetings/<id>/notes.md
abschrift process-dir ~/Abschrift/meetings/<id> --provider builtin
Good to know
- Live recording stays in the app. Capturing the mic and system audio needs macOS capture permissions. Once a recording exists on disk, everything after it is scriptable.
- The CLI is flag-driven. It deliberately doesn't read the app's settings. The
vault defaults to
~/Abschrift, and--outoverrides it. Models resolve from~/Abschrift/models, andABSCHRIFT_MODELS_DIRoverrides that. You choose the notes provider per invocation with--provider. Cloud providers read the key from theABSCHRIFT_API_KEYenvironment variable. - Same tiering as the app.
process,process-dirandimportcount toward the free tier's 5 meetings per month. The MCP server and--exportare Pro features. - The vault is plain files. Each meeting folder holds
notes.mdandtranscript.md. Reading them needs no CLI at all. The retention schedule applies no matter who wrote the meeting.