Privacy Policy.
Last updated November 19, 2026
Who we are
Orchestra Code ("Orchestra", "we", "us") runs AI coding agents on your own computer, in the Orchestra Code desktop app for Mac and Windows, and lets you follow and steer them from the web and from the mobile app for iOS and Android. This policy explains what we collect, why, and the choices you have. It applies to the desktop app, the mobile app, and orchestracode.com.
What we collect, and how
Account information. Collected when you create an account or sign in: your email address, a password you choose or the basic profile details (name, avatar) your sign-in provider returns, for example GitHub. We use a user identifier to keep accounts separate. Used to authenticate you, run the service, and contact you about it.
Your computers. Collected when you sign in to the desktop app: the name you give the computer and whether it is online. Used to list your computers in the app and to route your messages to the right one.
Your content. Collected when you start or continue a task: the prompts you type or dictate and the messages you send in the chat. The agent runs on your own computer and works on the code there, in your own Git repository. Our server keeps the task list (title, the request you started the task with, status, branch, and pull request link). For tasks that run on your computer, it does not keep your code or the conversation. Used to run the coding agent and to show you the chat, diff, preview, and pull request.
Images you attach. Collected only when you pick a photo for a task. We access your photo library at that moment and nowhere else. Used as part of that task.
Voice input. Collected when you use the microphone button. With a Streaming & sharing subscription, the app asks the first time whether your recording may be sent to our server. If you allow it, the recording is sent to our server, passed to our speech provider (Groq or OpenAI) to be turned into text, and discarded as soon as the text is back. We do not save audio recordings. The provider receives the recording and a few names from the task (repository, branch, file names) so they are spelled correctly, and no email address or account details. If you do not allow it, or have no subscription, your speech is transcribed by your device operating system speech service and no audio reaches us. Either way the text becomes your prompt. You can change your choice at any time in Settings, under Data and AI processing.
Credentials you connect. Collected when you add an AI provider key or connect GitHub or Bitbucket. Stored encrypted, never displayed back, and used only to run your tasks against your own accounts. Keys you set up in the desktop app stay on that computer.
Diagnostics. Collected automatically as your tasks run: timestamps, cost reported by your AI provider, and error logs. Used for security, debugging, and showing you what a task cost.
Purchases. Collected from Apple, Google, or Stripe when you buy a subscription: the purchase status and the product. Used to unlock your subscription. We never receive your full payment card details.
How we use your information
- To pass your messages to your computer and show you its tasks, diffs, previews, and pull requests.
- To operate your account and subscriptions.
- To keep the service secure and to debug and improve reliability.
- To respond to your support requests.
We do not sell your personal data, and we do not use your code to train our own models.
How the app reaches your computer
The agent runs on your computer, not on our servers. When you follow or steer a task from your phone or the web, your messages and the agent's progress travel between the app and your computer through our relay server, encrypted in transit. The relay keeps a task's live progress for at most one hour so the app can catch up, and does not store the conversation. When you open the code, the diff, or the terminal of a task, your computer sends that view through our server for that one request. The web preview of your own project is served from your computer through a Cloudflare tunnel.
Edit together. When you share a task with Edit together, the link and PIN unlock a space that is end-to-end encrypted. What you and the other person exchange is encrypted on your computers. Our server stores and passes on only encrypted data and cannot read it.
Third-party AI providers
What is sent. When you start or continue a task, the agent on your computer sends your prompt, your follow-up chat messages, any image you attached, and the files from the repository that it needs to read or change to an AI provider. Nothing is sent to a provider until you start a task.
Who it is sent to. Only the provider you connect, and only that one: Anthropic (Claude), OpenAI (Codex, the API, or your ChatGPT subscription), or OpenRouter, which routes to the open-weight model you pick. Orchestra is bring your own key: the request runs on your own account with that provider and is billed to you at their price. We do not add an AI provider of our own on top, and we do not send your content to any AI service you have not connected. The one exception is dictation you have allowed (see Voice input above): that recording is transcribed by Groq or OpenAI on our account, and is not saved by us.
Your permission. Before your first task, the app shows this disclosure and asks you to agree. You can read it again and withdraw your agreement at any time in Settings, under Data and AI processing. If you withdraw it, tasks stop running and nothing further is sent.
How they handle it. Each provider processes the request under its own terms. We only use providers whose terms commit to protection equivalent to this policy: processing the content to serve your request, not training their models on it by default, and applying encryption in transit and comparable security. Their terms are at anthropic.com, openai.com, and openrouter.ai. If you use your own key, the retention that applies is the one on your own provider account.
Git hosting
When you connect GitHub or Bitbucket, the agent on your computer uses that access to create branches, push commits, and open pull requests, and the app uses it to show and merge pull requests when you ask. We use the access you grant only to perform the tasks you start.
Data sharing
We share data only with the service providers needed to run Orchestra: Vercel (our website and API), DigitalOcean (our database and relay server), Cloudflare (the web preview tunnel from your computer), the AI provider you connected, the Git host you connected (GitHub or Bitbucket), and Apple, Google, or Stripe for purchases. Each acts under its own agreement, only to provide the service, and under terms we consider to give your data protection equivalent to this policy. We do not sell your personal data and we do not share it for advertising.
Data retention and deletion
We keep your account data and task list for as long as your account is active. Live progress on the relay expires after one hour. Your code and conversations stay on your own computer, under your control. You can delete your account at any time from Settings, which removes your account data. To request deletion or ask a privacy question, contact us at the email below.
Security
We use encryption in transit and at rest for sensitive data, including API keys. No system is perfectly secure, but we work to protect your information and limit access to it.
Children
Orchestra Code is not directed to children under 13, and we do not knowingly collect their data.
Changes to this policy
We may update this policy from time to time. We will revise the date at the top when we do.
Contact
Questions about this policy or your data? Email privacy@orchestracode.com.