A transcription CLI for AI agents and people. Turn audio and video files into text with timestamps or SRT/VTT subtitles, with speaker labels, in about 98 languages, through ScribeToAny. Coding agents (Claude Code, Codex, Cursor, Gemini CLI, OpenClaw) run it like any other command; the agent skill teaches them the options.
npm i -g scribetoany # or: npx scribetoany …
scribetoany auth login # opens the browser once
scribetoany transcribe meeting.m4a --format srt --speakersAgent skill: npx skills add scribetoany/skills --skill scribetoany.
| Command | |
|---|---|
auth login [--no-browser] |
Sign in with your ScribeToAny account (OAuth in the browser) |
auth logout |
Revoke the token and forget it |
auth status |
Show the signed-in account |
transcribe <file|url>… |
Upload (or import a public link), wait, save ./<name>.<format> |
get <id> [--wait] |
Save a finished transcript in any format |
list [--query q] [--limit n] |
Recent transcripts |
usage |
Plan, files today, minutes left this month |
Run scribetoany --help for every option. Results go to stdout, progress to
stderr; --json gives machine-readable output.
Exit codes: 0 ok, 2 bad input, 3 not signed in, 4 plan limit, 5 transcription failed, 6 network or server error, 7 not ready yet.
auth login opens ScribeToAny in your browser, where you sign in and allow
"ScribeToAny CLI" (OAuth with PKCE; the browser hands the result back to the
CLI on your own machine).
Tokens are stored in ~/.config/scribetoany/credentials.json (mode 0600;
%APPDATA%\scribetoany on Windows). The access token lasts an hour and is
refreshed automatically; the refresh token expires after 30 days unused.
Environment: SCRIBETOANY_BASE_URL (server), SCRIBETOANY_TOKEN (use this
access token instead of the stored one), SCRIBETOANY_CONFIG_DIR.