Privacy Policy
United States only for the hosted demo. The Demo allowance backed by our hosted keys is offered only to users in the United States. The Extension is globally available under the MIT license; if you’re outside the U.S., use My-Keys Mode or self-host — see the self-host guide. Peanut Gallery is free; there is no paid tier and no payment processing.
Short answer. Audio is streamed through a transcription provider and discarded — never written to disk, never used to train AI models. Your live transcript sits in server memory during the session and is not persisted. If you use your own API keys they travel per-request as headers and are never stored server-side. We don’t sell data. The Extension loads no analytics library. The marketing site uses Cloudflare Web Analytics (cookieless, no cross-site tracking, no individual identifiers) for aggregate page views and Core Web Vitals.
Recording-consent laws are your responsibility. The Extension can be pointed at any browser tab you grant it. Many U.S. states require all-party consent to record private conversations.
Peanut Gallery is a Chrome extension, a hosted backend at api.peanutgallery.live, and a marketing site at www.peanutgallery.live. This policy describes what data the Service handles, what we store, and for how long. It reads alongside the Terms of Service. The Service is provided by Manu Games LLC, a North Carolina limited liability company.
How you can use the Service today
You pick at runtime in the Extension’s Backend & keys drawer:
- Demo Mode — 15 minutes of free transcription on our API keys, one-time per install. Gated by a per-installation UUID (
installId) generated locally. - My-Keys Mode (BYOK) — you supply your own API keys. We forward them per-request to the providers and never store them server-side. In this mode, Manu Games is not a processor on the AI-inference path: you are in direct privity with the provider whose key you supplied, under that provider’s own terms and privacy policy.
- Self-host — deploy the MIT-licensed backend yourself. We don’t see any data in this mode; this policy doesn’t apply to your self-hosted instance.
No paid tier. Peanut Gallery is free. There is no subscription, no payment processing, and no billing code in the product. Earlier versions of the repository contained scaffolding for a planned “Peanut Gallery Plus” tier; it has been removed and is not offered or planned. We collect no email address and no account.
What the Extension keeps on your device
The Extension writes to chrome.storage.local, which Chrome persists on your disk. Nothing there leaves your browser unless you trigger it:
- Your chosen backend URL, pasted API keys (if any), persona pack, theme, response-rate dial, sensitivity level, and per-persona mute list.
- Your per-install UUID (
installId) — used to meter the Demo trial. - Up to 50 prior sessions’ metadata, sampled transcript tails (capped at 4,000 characters per session), and persona reactions — groundwork for a future “Past sessions” feature. LRU-evicted at 50.
- Per-session upvotes / downvotes / pinned entry, plus UI state.
What the backend receives and keeps
The hosted backend at api.peanutgallery.live is a stateless API behind a session ID. When you stream a session it handles:
- Audio — received as PCM, forwarded to the transcription provider with the
mip_opt_out=trueflag set, and discarded. Nothing is written to disk. The opt-out flag tells Deepgram not to retain or use your audio to train AI models. - Live transcript — held in memory during the session, routed to the Director and personas, discarded on session end.
- Your API keys (My-Keys Mode) — forwarded per-request to the providers, redacted from every log line, never persisted.
- Your install ID (Demo Mode) — kept in memory to meter the 15-minute trial.
- Operational logs — request IDs, session IDs, persona IDs, response times, error events. 30-day retention. Redacted of keys; never include complete transcripts or audio.
We persist no account, email address, or payment information — there is nothing to buy and no sign-up.
AI training on your content
We do not train AI models on your audio, transcripts, or content, and we configure each provider to minimize retention to the extent each provider offers that control:
- Deepgram —
mip_opt_out=trueis sent on every transcription request. Under Deepgram’s Model Improvement Partnership Program documentation, this limits retention to the duration necessary to process the request and excludes the content from model training. - Anthropic — API traffic is non-training by default per Anthropic’s commercial posture. We do not opt in to any training-partnership program.
- xAI — Enterprise ToS §3.3 provides that xAI shall not use User Content for training and auto-deletes within 30 days absent a legal hold or safety flag. This covers both the persona-generation calls AND the fact-check search queries, which run through Grok Live Search.
These configurations protect content flowing through the Service. They do not control what you do with persona output after it leaves the Service, or what providers do under their own terms with your account in BYOK mode.
Feedback on persona reactions
When you click the menu on a persona reaction and choose upvote, downvote, pin, or quote-card, we log that action to /api/feedback. Each log entry contains the action, the persona, the pack, the reaction text, the transcript tail (~500 characters) that informed the reaction, your install ID, and a timestamp. We retain these for 90 days.
We use them to improve the personas’ prompts, the Director’s routing, and the session-recall highlight picker. Feedback is keyed to the pseudonymous install ID only — we don’t tie votes to your email or license key.
Opting out: today, either self-host the backend with DISABLE_FEEDBACK_LOGGING=true or don’t interact with the feed menu. A user-facing opt-out toggle in the Extension is on the roadmap and should ship before the first public paid launch.
Third-party providers
When the Service calls a third party on your behalf, that provider’s terms and privacy practices apply. Current subprocessors:
- Deepgram — speech-to-text streaming.
- Anthropic (Claude) — Producer and Joker persona inference.
- xAI (Grok) — Troll and Sound FX persona inference, and fact-check search queries via Grok Live Search.
- Railway — backend hosting.
- Vercel — Next.js application hosting where used.
- GitHub Pages — marketing-site hosting.
- Cloudflare Web Analytics — cookieless, privacy-respecting page-view metrics on the marketing site only (not in the Extension). No cookies, no cross-site tracking, no individual identifiers.
Marketing-site analytics
The marketing site at www.peanutgallery.live loads Cloudflare Web Analytics to measure aggregate page views and Core Web Vitals. It is cookieless and does not use any client-side persistent identifier; it does not track users across sites. Cloudflare receives your IP address, user agent, referring URL, and page-view events, and uses them only to compute aggregate metrics. We do not run any other analytics or advertising libraries on the marketing site. The Extension itself does not load any analytics library.
Retention summary
| Data | Retention |
|---|---|
| Audio bytes | In-memory, during session only — mip_opt_out=true sent to Deepgram on every request |
| Live transcript | In-memory, during session only |
| Voted / pinned / quote-carded snippets | 90 days |
| API keys (BYOK) | Never stored |
| Install ID | Backend process lifetime |
| Operational logs | 30 days |
| Marketing-site analytics (Cloudflare Web Analytics) | Aggregated metrics; no per-visitor identifier stored |
Your rights
Wherever you live, you have the right to know what we hold, correct it, delete it, port it, object to certain processing, and complain to a data-protection authority. Under U.S. state privacy laws (CCPA/CPRA, Virginia, Colorado, Connecticut, and equivalents) you also have the right to opt out of “sale” and “sharing” of personal information. We do not sell or share personal information in the senses those laws define.
To exercise any of these rights, email [email protected]. Because we collect no email or account, the only data tied to you is pseudonymous (your install / session IDs and any feedback snippets). We respond within the timeframes required by law — typically 45 days under U.S. state privacy laws and one month under GDPR.
International data transfers
The hosted backend is in the United States. If you use the Service from outside the U.S., your audio and transcript pass transiently through U.S.-based infrastructure and are not retained. Where applicable, we rely on the European Commission’s Standard Contractual Clauses, the U.K. International Data Transfer Addendum, and (where participating) the EU–U.S. Data Privacy Framework. If that’s not acceptable for your situation, use My-Keys Mode or self-host the backend in a region of your choice.
Children
Peanut Gallery is not directed at children under 13 (16 in some EU member states). We do not knowingly collect personal information from children under those ages. If you’re a parent with a concern, email [email protected] and we’ll help.
Security
HTTPS for all backend traffic, keys redacted from logs, structured per-request timeouts on every upstream call, narrow content-script scope (peanutgallery.live subdomains only). Responsible disclosure via SECURITY.md or [email protected].
Recording consent
The Extension can be pointed at any browser tab you grant access to. Recording-consent laws are jurisdiction-specific. U.S. states including California, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Nevada, Pennsylvania, and Washington require all-party consent for private conversations. The EU, UK, and many other jurisdictions impose equivalent or stricter rules. You are solely responsible for obtaining any consent required before directing the Extension at audio — this applies to live calls, conferences, voicemail, ambient room audio, and private streams. See the Acceptable Use section in the Terms of Service.
Changes to this policy
Changes are effective on the updated “Last updated” date above. If we ever introduce a material new data flow (for example, accounts or a paid tier — not currently offered or planned), we’ll describe it here before it takes effect.
Contact
- Privacy questions and data-subject requests: [email protected]
- Legal, billing, DMCA: [email protected]
- Security reports: [email protected]