- file
- open-core.transcript
- speakers
- You, vocateca
- address
- vocateca.com/en/open-core
- duration
- 00:04:12 · 589 words
The part that touches your audio is open
What is open
Partly, and deliberately so. Vocateca is open core: the CLI, the MCP server, and the on-device transcription engines — the code that actually touches your audio — are open source under the Apache-2.0 license. The macOS app itself, and the account and billing backend behind Pro, are proprietary.
The open-core repository contains the entire transcription pipeline: everything that downloads, transcribes, organizes, and exports your audio. There is no entitlement check anywhere in it — every feature works with no account and no Pro gate.
VocatecaCore | The domain layer: subscription and source management, the download-and-transcription pipeline, local state (SQLite), speaker diarization, proper-noun correction, and integrations (webhooks, Notion, Obsidian export). Headless, no UI. WhisperKit (Whisper) ships inside it as the reserve engine, for languages outside Parakeet-TDT’s 25. |
VocatecaParakeet | The Parakeet-TDT engine binding: Core ML on the Apple Neural Engine. The default engine. |
VocatecaQwen | The Qwen3-ASR engine binding: MLX on Apple silicon. Optional; you choose it in Settings, and it is never picked automatically. |
vocateca-cli | A scriptable command-line interface over the core, with --json output on every command, and a built-in MCP server (vocateca-cli mcp) so an AI assistant can drive it directly over stdio. |
| Mac app | The interface and the Pro automation process. |
| Account and billing | Sign-in, checkout and entitlement. |
Why
You can verify the privacy claims
"Nothing leaves your Mac" is easy to say in marketing copy. In the open core it's something you can actually check — read the exact code that downloads, transcribes, and stores your audio, and confirm for yourself that there's no phone-home and no telemetry.
No lock-in
Transcripts are plain Markdown, SRT, and OKF files on your disk, not a proprietary format trapped in an app. The CLI and MCP server carry no entitlement check, so even if you stopped using the macOS app tomorrow, the engine underneath keeps working exactly as it does today.
Funding that doesn't compromise you
Vocateca isn't funded by ads, by selling usage data, or by a free tier designed to convert into data collection. It's funded by people who choose to pay for the polished app and Pro automation — a business model that only works if the underlying claims are true, which is the whole point of opening the engine.
What stays closed
The macOS app's interface, its scheduled automation daemon, and the account, checkout, and entitlement backend behind Pro are proprietary. None of that code is about how transcription happens — it's the packaged product experience and the billing system that funds building it. Keeping it closed is what lets the engine underneath stay free and genuinely unencumbered, with no paywall hidden inside it.
- The macOS app UI and the Pro automation daemon (scheduled auto-download, background transcription, daily summaries)
- The account, checkout, and entitlement/billing backend
What this means for you
Audit it
Read exactly how your audio is downloaded, transcribed, diarized, and stored. Confirm for yourself that there's no network call this page didn't already tell you about.
Script it
Run vocateca-cli headlessly on a server, in a cron job, or from your own tooling. Every command supports --json output, and nothing in it is gated behind a subscription.
Build on it
Point an AI assistant at the MCP server, or embed VocatecaCore, VocatecaParakeet, or VocatecaQwen as libraries in your own project. Apache-2.0 permits commercial use, so you can build on it, not just around it.