“Private” needs a concrete definition
Call-coaching products often describe themselves as private, local, or user-first. Those labels are not enough. A buyer should be able to trace what happens to audio, transcript text, generated outputs, account metadata, and shared calls.
Ask five questions:
- Where is audio processed?
- Is raw audio saved?
- Where is transcript text stored?
- Who can access or share it?
- Is any data used to improve models or prompts?
User control during the call
A coaching tool should make its active state clear. Users should choose when a call begins, which live-assistance features are enabled, and when a card is dismissed.
Cinder’s call model has separate layers. On-Demand analysis is available on every plan and appears when the user requests it. Continuous Coaching is available on Starter and up. Pro adds Script Assist. Coaching and Script Assist can be enabled independently on Pro.
Cinder’s Free data flow
On Free, transcription runs locally on the Mac. Raw audio is neither uploaded nor saved. Transcript text is stored in the user’s Cinder account for the current Free access window shown on Pricing.
Free-plan transcript text may be used to improve Cinder’s AI under the disclosure accepted in the app. Identifying speaker names are removed before eligible text enters that process, and the data is associated with a pseudonymous identifier. Withdrawing the acknowledgement triggers a best-effort removal request rather than a promise of instantaneous deletion from every prior processing stage.
Cinder’s Starter-trial and paid data flow
During the default Starter trial and on paid Starter or Pro, live audio is streamed to a cloud transcription provider and is not saved as audio. Trial transcript text stays local-only. On paid Starter and Pro subscriptions, transcript text is stored on the Mac and in the user’s Cinder account by default.
Paid users can disable cloud transcript storage. Cinder then removes cloud transcript text and keeps future transcript text local, while account metadata and generated outputs such as summaries may remain. Paid transcript access is unlimited for transcripts the user retains; current automatic-cleanup defaults and available retention windows are shown on Pricing.
When coaching or analysis is requested, relevant conversation context is sent to Cinder’s configured AI providers. Paid-plan data is not used to train AI models.
Sharing is a separate decision
Local storage, cloud storage, and sharing are different controls. If a user shares a call with a teammate or through a share link, the recipient can access the shared transcript. Teams should decide who may share, what should be redacted, and how long the shared material is needed.
Connected integrations create another boundary. The currently configured providers, action tiers, one-click support, and automation support are shown on Pricing. Review authorization and destination before sending call-derived content to another system.
Bot-free does not mean undisclosed
Cinder captures Mac audio without adding a participant bot. That reduces meeting-platform dependency, but it does not remove laws or workplace policies about recording, transcription, monitoring, notice, or consent.
The person using Cinder remains responsible for giving any required notice and obtaining any required consent. A trustworthy workflow should make that responsibility explicit instead of using bot absence as a promise that other participants will not know.
A procurement checklist
- Read the Free and paid data flows separately.
- Confirm the default, not only the available setting.
- Test transcript deletion and retention controls.
- Review who can share calls and run connected actions.
- Ask which providers receive audio or conversation context.
- Document the notice and consent process for the jurisdictions and workplaces involved.
- Avoid compliance or security conclusions that are not supported by current evidence.
See Cinder’s Privacy Policy, Terms of Service, and plan comparison.