Third-party notices
Last updated: 29 September 2026
Cueshot is built on the work of others. This page names it; the full license texts and copyright notices are in the app itself, under Settings → Updates → Open-source licenses (a file that also opens without a network). Third-party names and marks belong to their owners and are used only to say what the software is and where it comes from — none of them endorses Cueshot.
Speech models
| Model | Author | License |
|---|---|---|
| Parakeet TDT 0.6B v3 — speech recognition on your Mac, in a Core ML conversion by FluidInference. Modified (converted and optimised for Apple silicon). | NVIDIA | CC-BY-4.0 — model card |
| Speaker recognition for meeting notes — segmentation and speaker embeddings from pyannote.audio and WeSpeaker, in a Core ML conversion by FluidInference. Modified. | Hervé Bredin et al. (pyannote); the WeSpeaker authors | As stated on the model page |
Software in the app
| Component | License |
|---|---|
| FluidAudio (runs the models), with NVIDIA NeMo text processing, rustfst, flate2, fastcluster and VBx | Apache-2.0 (fastcluster: BSD-2-Clause) |
| Sparkle (updates), with bsdiff, sais-lite, ed25519 and SUSignatureVerifier | MIT and BSD-style, as set out in its license file |
| The MCP server that hands captures to your agent — bundled with the MCP TypeScript SDK, Express, Hono, zod, ajv and about 85 further packages | MIT, ISC, BSD-2-Clause, BSD-3-Clause |
Fonts
Mona Sans (GitHub), Martian Mono (Evil Martians) and Geist (Vercel) are licensed under the SIL Open Font License 1.1; the license texts ship with the font files in the app.
Legal notice · Privacy · Terms