Privacy policy
Updated October 28, 2026
Lyre runs your projects on your computer. You do not need an account to use Lyre on your own network, and you do not need to upload your project to a Lyre-hosted builder. If you sign in and use the encrypted relay, Lyre AI or another hosted feature, the account, directory and usage data described below apply. Relay traffic is end-to-end encrypted between your host and authorized client.
Lyre is operated by Lyre LLC, a Texas limited liability company based in the United States. Lyre LLC is the controller of the personal information this notice describes. Before October 26, 2026, Lyre was operated by Jonathan Bakhit, an individual in Texas. Mailing address: 11214 Bandit Trail Dr, Houston, TX 77095-7153, USA. For privacy questions or requests concerning information we hold, contact support@lyrestudio.net.
This notice covers Lyre's applications, website and services. Optional features have different recipients and settings. A feature described here may be unavailable in your version, region or plan; describing it does not enable it or authorize a new transfer.
Local project access
Local previews run on your computer. Projects do not need to be uploaded to a hosted builder. You select the project folders Lyre can access.
Lyre can store working data on your devices, including project information, unsaved drafts, checkpoints and diagnostic logs. These local records are distinct from uploading your project to a hosted service. Optional provider integrations and services you choose to use may handle data separately, as described below.
Local conversations, attachments, screenshots, tool results and diagnostics may contain project text or personal information. Backups or synchronization tools you configure may copy those local records. Review what you share with an agent, another device, support or an external service, and obtain any permission required for other people's information.
Account and host directory
Signing in is optional. If you sign in, account and authentication data is handled by Supabase, our account and database provider. This includes the email address on your account, your account identifier, your sign-in method (email, Google, GitHub or Apple, where available) and the session records that keep you signed in. We also keep host records and billing state associated with the account.
While a computer is signed in to your account, Lyre publishes a short-lived directory entry so your own devices can find it. Each entry contains the host's server identifier, its public key, the relay endpoint and whether TLS is used, plus two display-only fields: the computer name (your operating system's hostname) and the operating system (macOS, Windows or Linux). The entry is stored against your account and is only ever listed back to your account. It is not shown to other users, and it is not an access grant. Entries expire, and they are removed when you sign out, withdraw a host, or when the lease ends.
Optional AI providers
Optional agents may send prompts, project context and tool output to their providers under your account settings. Lyre does not yet enforce exact outbound-context approval for CLI agents. Your chosen provider's terms, privacy policy, and account settings apply to information sent to that provider.
Choosing a local model does not prevent a CLI agent from making network requests or sending data to external services. CLI file and network access follows the provider and your environment settings; local inference is not a network isolation guarantee.
Your selected connection may use your own API key, a native provider sign-in, a local model or a routing gateway. A gateway can receive the request and pass it to another provider. Automatic switching or fallback can send conversation context to the next configured connection. Review the recipients and their terms before enabling a route; switching does not recall information already sent. Connecting your own account does not mean every provider has the same training, retention or regional-processing policy.
Managed Lyre AI
Lyre AI is an optional paid plan with an included monthly usage allowance. When you use it, your host obtains a short-lived, host-scoped key from our server and sends model requests to OpenRouter, which routes them to an inference provider. Requests include your messages and conversation history, agent instructions, and the project text, file lists, search results and tool arguments or results the agent uses. We record metered usage for your account: the amount reserved and settled for each request, the host and session it came from, and the running total against your monthly allowance. We do not record the prompts you send or the replies you receive in those account metering records. Local conversation history, optional diagnostics and reports you submit are separate records.
If an app version offers a limited hosted model option without a paid plan, your host sends those requests through a Lyre server function (Supabase) to OpenRouter, which routes them to an inference provider. The server function screens the request and reply and may refuse a turn it flags or cannot screen. We keep a per-account daily count to apply usage limits and do not record your prompts or replies in that count.
Lyre AI uses OpenRouter to route requests, and DeepInfra is the inference provider for the current managed models. For paid Lyre AI, safety screening on your computer is optional and off unless you turn it on under Connections > Routing. If you turn it on, screening checks your messages, agent instructions, candidate replies and supported reasoning, and the project text, tool arguments and tool results used in the turn, by sending them to a screening model hosted on DeepInfra through OpenRouter. Screening is advisory: it can flag or log content but does not by itself stop your work, and it is not a promise that secrets or personal information will be removed. Screening usage counts against your Lyre AI allowance. Turning screening off stops further screening requests; it does not recall earlier ones.
For these managed requests Lyre asks OpenRouter to use only its disclosed DeepInfra endpoints, to request zero data retention, to deny provider data collection and to disable recipient fallback. These are request controls, not a promise that every Lyre service, account record or independently configured provider has zero retention. OpenRouter and the inference provider have their own processing terms. See OpenRouter's data controls and DeepInfra's privacy information.
Only include project or workplace information you are permitted to share with these services. Do not include passwords, API secrets, payment details, health records or other restricted sensitive data in managed AI requests unless a separate arrangement expressly permits that processing and its required safeguards. Safety screening is not a promise that secrets or personal information will be removed before content reaches a provider.
Lyre Cloud
Where Lyre Cloud is offered to your account, it runs a computer for you on Lyre's cloud infrastructure instead of your own machine. That computer holds the project files, conversations, running processes and storage you put there, so Lyre and its infrastructure provider can technically access them as the operator of that computer. Cloud uses the same account and, for remote viewing, the same relay as your other computers. It is separate from projects that stay on your own computer, which are not uploaded. Cloud computers and their storage are tied to your account.
Referrals and backer rewards
If you use a referral link or code, or receive a backer reward, we record the code, the accounts it links and the resulting discount or credit so the offer can be applied once. Payment for referred purchases is handled by Stripe. We do not read your contacts, and a referral does not share your project content.
Shared conversations
Where conversation sharing is offered and you share a conversation, the snapshot you review, with its title and the content in it, is stored on Lyre's servers (Supabase) so anyone with the link can open it. A share is limited to 2 MiB and expires after at most seven days, and you can revoke it sooner, which removes the stored snapshot. A conversation can contain source, file paths, screenshots and other people's information, so review it before sharing. Copies can remain in Supabase's rolling database backups for up to seven days after removal.
Push notifications
If you enable push notifications on a phone, your host sends the notification title, text and routing data to Expo's push service, which delivers it through Apple or Google. That text is not covered by relay end-to-end encryption and is visible to those providers and on your lock screen. You can turn notifications off in the app.
Voice input
Voice input is optional. Where managed voice is offered and you choose it, your host sends recorded microphone audio through Lyre's managed service, hosted with Supabase, to Groq for transcription. The service and Groq process that audio; the returned transcript can become part of your local conversation and can be sent to the AI provider you use for that conversation. Usage and entitlement records are separate from the recording.
Groq's terms and retention controls apply to its processing. Its data guide describes limited reliability or abuse-monitoring retention and optional zero-data-retention controls; this notice does not promise that every account has those controls enabled. See Groq's data handling information. If you choose a local speech engine or configure another speech provider, that choice has a different processing path. An operating-system microphone permission does not mean the audio stays on your device.
Payments
Web subscriptions are processed by Stripe on Stripe-hosted pages. Card details are entered on Stripe's pages and are never received or stored by Lyre. We store the resulting customer identifier, subscription status and entitlement. Purchases made inside the iPhone or Android app are processed by Apple or Google; we receive purchase and transaction identifiers to verify and maintain your subscription.
Device connections
Direct connections on your private network are preferred. Those local-network connections may use unencrypted HTTP, so treat a local network as trusted. For supported relay connections, your host establishes an outbound connection, and project traffic is end-to-end encrypted between the host and authorized client. The relay, hosted on Cloudflare, forwards encrypted payloads without the keys needed to read their contents. Access is scoped to the grants issued by your host.
Encryption depends on the path, not on the app as a whole. Relay project traffic is end-to-end encrypted. Requests to our cloud services (account, billing, voice and hosted features) use HTTPS. Some direct local-network connections use unencrypted HTTP, and traffic that reaches an optional AI provider you configure is protected only by that provider's own transport.
Encryption protects the contents of relay traffic; it does not hide all connection metadata. The relay can see IP addresses, connection timing, message sizes, session identifiers and the public handshake keys used to set up encryption. It cannot read your project files, prompts or messages. Web and relay hosting run on Cloudflare, which may keep limited operational logs.
Remote desktop and app previews
When you view or control a computer remotely, or preview an app on another device, the stream shows whatever is on the screen or in the app, including notifications, passwords and other people's information. On a relay or fallback path Lyre's infrastructure can see connection metadata such as timing and volume. Only share a screen you are allowed to share and tell people whose information may appear. The operating system may ask you to allow screen capture, and sharing an app is not permission to view the desktop.
We send transactional email, such as sign-in links and account notices, through our email provider, Resend.
Workplace integrations, tools and publishing
If you connect Slack, Discord, Notion or Linear, Lyre can use that connection to retrieve the messages, threads, pages or issues you select. Those services receive the API request and relevant account or content identifiers. Content added to an agent's context may then be sent to that agent's AI provider. Only connect information you are permitted to use this way, including under your workplace's policies.
Plugins, MCP servers, terminal commands, browser tools and applications you preview can contact their configured services. Optional publishing or deployment sends the selected project or build to the destination you choose. Those destinations' access, retention and privacy settings also apply. Permissions to view a shared app do not by themselves grant access to your agent, project source or whole desktop.
Downloads and routing diagnostics
Checking for updates, browsing Marketplace or downloading templates, agent tools and local models contacts the relevant distribution host, such as GitHub, GitHub Pages, npm or the model publisher. The host receives ordinary connection and request information, including your IP address and the resource requested. An installed template or tool can have its own dependencies, network behavior and license.
A routing gateway may keep local or operator-managed request histories, caches and diagnostic exports, potentially including prompts, replies or project context. Its logging settings are separate from Lyre account usage records and from an upstream provider's settings. Review the gateway's operator and settings before sending sensitive information. Share diagnostic exports with support only after reviewing their contents.
Hosted safety reports and contributions
Where Lyre's hosted AI-response or shared-app safety reporting is available, a report can include the response excerpt you choose, your explanation and any contact information you supply. Older builds may not offer the reporting action. Reports are stored separately from encrypted relay traffic, and Lyre's operator can read the content you deliberately submit.
The current hosted reporting policy deletes the live report when review is resolved and caps unreviewed reports at 30 days, with scheduled pruning. Earlier copies can remain in Supabase's rolling daily database backups, currently available for seven days; a restored database must be pruned before it is returned to service. Reviewer exports or support correspondence are separate records and must also be handled under the retention process. Disabling reporting does not itself erase an earlier report or backup. Contact support@lyrestudio.net for questions about a report you submitted.
This section describes reports sent to Lyre's hosted safety-reporting service. If you prepare a local app-run bug-report draft or diagnostic export, it is separate and can contain screenshots, notes and app context. Creating a local draft does not upload it to this hosted receiver. Copying, saving or manually sharing a report is a separate action; review the included information and the destination you choose.
Optional Sponsor contributions are one-time payments made through a Stripe-hosted page. Stripe processes them under its privacy policy. Sponsor payments are separate from Pro subscriptions and grant no product features or benefits.
Website analytics
Our public website is hosted by Vercel and uses Vercel Web Analytics and Cloudflare Web Analytics to measure visits, page views and website performance. These services process page addresses and referrers, approximate location, browser and device information, and performance measurements to provide traffic and reliability statistics. These analytics products do not use analytics cookies. Our Vercel integration removes URL query strings and fragments and excludes account and Slack connection pages. Cloudflare has its own collection rules; its documentation says it does not log query strings. Website analytics is separate from Lyre's agent conversations and local project processing. See Vercel's analytics privacy information. See also Cloudflare's Web Analytics information.
Service providers
We use these providers only for the purposes described above: Supabase (accounts, database, server functions and shared-conversation storage), Stripe (web payments, referral credits and Sponsor contributions), Apple and Google (in-app purchases and push notification delivery), Expo (push notification delivery), Cloudflare (relay, app hosting and public website analytics), Resend (transactional email), Groq (managed voice transcription), OpenRouter (managed AI routing), DeepInfra (managed AI replies and optional screening) and Vercel (public website hosting and analytics). Lyre Cloud also uses a cloud infrastructure provider to run Cloud computers. Your optional provider, gateway, workplace, tool and deployment choices can add the recipients described in their sections. We do not sell your personal data.
Purposes and international processing
We use information to provide the features you request, authenticate accounts and devices, maintain allowances and purchases, deliver service messages, handle reports and support, secure the service and meet legal obligations. Where applicable privacy law requires a legal basis, these activities rely on performing our agreement with you, applicable legal obligations, and legitimate interests such as service security and support. Where a feature or applicable law requires consent, we ask for it and provide a way to withdraw it.
Lyre's operator is based in the United States. Cloud providers may process information in the United States or other countries under their service arrangements. A local model, encrypted relay or zero-retention request does not establish that all processing occurs in your country. Applicable transfer safeguards and service restrictions depend on the recipient and processing involved. Contact us for information about the safeguards that apply to your data; we do not claim that a US operator or a generic opt-in alone satisfies every international-transfer requirement.
If you are in the European Economic Area or the United Kingdom, your information is processed in the United States and wherever our providers operate. Transfer safeguards depend on the provider, and you can ask us which apply to your data. You also have the right to complain to your local data-protection authority. Contact support@lyrestudio.net for any request.
Your control
You can stop local previews and manage paired-device access from your host computer. Local source files remain editable outside Lyre. Stopping a preview does not remove information already sent to an optional AI provider.
Account deletion
To request deletion of your Lyre account and associated data, delete it from the account screen in the app, follow the account deletion steps or email support@lyrestudio.net. Requests are reviewed in about two days. After ownership verification, we aim to remove the account and its associated live data, including account-linked directory entries, host associations, shared-conversation snapshots, referral records, account text and Lyre-held billing records, within one week. Backup handling is described below. Deletion is complete once the required removal is verified and deletion requests have been sent to the service providers that processed the account's data; those providers act on requests under their own processes.
After deletion is verified complete, Lyre keeps only two records: a hash of the deletion receipt with its completed status for 24 hours, so the requesting device can confirm the result, and a permanent fingerprint of each app-store purchase source (the store and a SHA-256 hash of its purchase identifier) so an old purchase cannot be claimed by another account. The fingerprint contains no account, computer or contact details, but someone who holds the original purchase identifier could match it to that purchase. Supabase keeps project-level database backups on its own schedule, and Stripe, Apple and Google may keep payment and purchase history as the law and their own policies require. Deletion stops a Stripe subscription from renewing; it does not refund the current period or cancel an Apple or Google subscription. The guide explains local files and independent provider accounts, which this request does not erase.
Retention, privacy rights and changes
Local records remain on your devices until you remove them using the relevant controls or local data process. Hosted records are retained for the feature's purpose, the deletion and reporting rules above, and any applicable legal requirement. Provider retention and backup schedules are separate; stopping a feature or withdrawing consent does not recall information already processed. We do not promise a single retention period for all data.
Depending on applicable law, you may ask to access, correct, delete or receive a copy of personal information, restrict or object to processing, or withdraw consent. Contact support@lyrestudio.net; we may need proportionate information to verify your request. We respond to verified requests within the time applicable law allows, generally one month, and explain any lawful limitation. You may also complain to the relevant data-protection authority. This policy does not limit rights that applicable law gives you.
We update this notice when our practices change and identify the revision date above. Where required, we provide an additional notice or request renewed permission before a material new use or transfer. A published notice does not by itself grant permission to send your data to a new optional service.
Children
Lyre is intended for adults aged 18 and over and is not directed at children. Lyre does not verify age on every platform; on iPhone and iPad it uses Apple's age signals where Apple requires them. If you believe a child has provided personal information, contact support@lyrestudio.net and we will remove it.