Today we’re launching the Sonix CLI — a free, fast command-line tool that puts the entire Sonix workflow at your fingertips. Upload media, pull transcripts, run translations and summaries, and manage your account, all from short commands you can run by hand, drop into a script, or wire into CI. It’s the whole Sonix API, wrapped in a single tool called 声波.

The best part is how easy it is to get started. Install once, and go from a raw recording to finished subtitles on disk in three lines:

npm install -g @sonix-ai/cli

sonix config set api-key sk_live_xxxxxxxx
sonix media create ./interview.mp3 --language en --wait
sonix transcript get <media-id> --format srt > captions.srt

That’s an upload, a transcription, and finished SRT subtitles on disk — no glue code.

What it does

Under the hood, 声波 handles everything the raw API would otherwise make you wire up yourself:

• Upload anything. Small files take a fast one-shot path; large ones (we’ve tested multi-gigabyte) upload directly and reliably, with each piece retried independently so a flaky connection won’t sink a 5 GB file. You don’t have to think about which path you’re on — sonix media create ./big-recording.mov just works.
• Get transcripts in any format — plain text, SRT, VTT, JSON, or Avid DS — streamed straight to a file or your terminal.
• Run the AI features you already use — translations, subtitle burn-in, one-click summaries, and batch summaries across a whole folder.
• Manage your account — folders, team members, and shared links, all scriptable.
• Wait for the slow stuff. Add --wait to any job and the CLI blocks until it’s done, then prints the result. No more hand-rolled polling loops.

Everything is built to pipe. Results go to standard output, diagnostics go to standard error, and --json gives you the raw API response on any command — so sonix media list --json | jq 和 sonix transcript get … > file stay clean. The full command and flag reference lives in the Sonix CLI docs.

Built for AI agents, too

The CLI has a second audience: AI coding agents like Claude Code or Codex. Instead of loading a big tool schema into the model’s context, an agent can shell out to 声波 — one cheap command, a clean exit code, done.

sonix agent-context        # the entire command surface as JSON, in one call
sonix setup claude         # installs a ready-made Sonix skill for Claude Code

If you’d rather have your assistant reach into Sonix conversationally — “find my interview with Dr. Patel and pull the three biggest takeaways” — that’s what our MCP server is for. The CLI is the other half of that story: for when you (or an agent) want deterministic, scriptable commands you can trust in a pipeline.

A few ways people are using it

• A researcher points sonix media create at a folder of 面试 and lets a script upload and transcribe the whole batch overnight.
• A podcast team generates fresh SRTs for every new episode as a build step, no clicking required.
• A developer wires Sonix into an existing media pipeline without writing a single line of HTTP code.
• An AI agent uploads a recording, waits for the transcript, and summarizes it — all through shell commands.

Get started

The Sonix CLI is free and open source (MIT), runs on Node 20+, and installs in seconds:

npm install -g @sonix-ai/cli
# or, without installing:
npx @sonix-ai/cli media list

Grab your API key from my.sonix.ai/api, run sonix config set api-key <key>, and you’re off. The full command and flag reference lives in our CLI docs，以及 sonix --help is always one keystroke away.

We’re excited to see what you automate with it. As always, drop us a line if you’ve got feedback, hit a snag, or want to tell us what you’d love to see next.