Starlight Search Inc. ("Starlight," "we," "our," or "us") operates Pulsar, on the web and in the Pulsar iOS app (together, the "Service"). This Privacy Policy explains what information we collect, how we use it, who we share it with, and the rights you have over it.
1. Information we collect
Account information
When you sign up, our authentication provider (WorkOS AuthKit) collects your name, email address, and a hashed credential or the identifier of the account you sign in with (Google or Apple). If you use Sign in with Apple and choose to hide your email, we receive an Apple relay address instead. We store a corresponding user record in our database along with the organization you belong to.
Workspace content
Your chats with the agent, the files and images you upload, and the files the agent writes on your behalf are stored in isolated per-user storage, encrypted at rest, in the United States. Pages you publish are stored so the people you share them with can open them. We do not train models on your workspace content.
Optional features
- Voice input: when you dictate, your audio is streamed directly to our speech-to-text provider (Deepgram) to produce the text. Pulsar does not store the audio.
- Notifications: if you allow notifications, we store your device's push token (Apple Push Notification service, or your browser's push service) to tell you when the agent finishes or needs you.
- Connected apps: when you connect an app (for example Gmail, Notion or Linear, through Composio or the app's own MCP server), the agent reads and acts on that app's data at your request. Access tokens are held encrypted.
- Agent web browsing: the agent can use a cloud browser (Steel). Sign-ins you complete in it are saved for that project so the agent can return to the site.
- Bug reports: when you report a bug you can attach the conversation; it is sent to our issue tracker (Linear) so we can fix the problem.
Usage data
We log API requests, sandbox lifecycle events, model-usage cost, and error traces to operate, secure, and bill the Service. These logs are retained for up to 90 days. We also record product analytics (pages viewed, features used, and error reports) linked to your account, through PostHog. We do not use your data for third-party advertising and we do not sell it or share it with data brokers.
Billing information
Payments are processed by Stripe. We receive a subscription status and customer identifier. We do not store your full card details.
2. How we use information
- Provide, maintain, and improve the Service.
- Enforce plan limits (daily model-usage budget, project counts, storage caps).
- Detect abuse, fraud, and security incidents.
- Communicate with you about your account, incidents, and material changes.
- Occasionally email you about unfinished work after you have been away; you can turn this off in Settings → Notifications or from the link in any such email.
- Comply with applicable legal obligations.
3. Model providers and subprocessors
When you use the Service's AI models, your prompts and relevant workspace context are transmitted to the model provider for inference. Most models are reached through OpenRouter, which passes the request to the model's provider (for example OpenAI, Anthropic, Meta or DeepSeek); some deployments also use Amazon Bedrock. We use these providers under terms that do not permit training on your content.
Other subprocessors include:
- Railway: application hosting.
- Neon: managed Postgres database.
- Fly.io: sandbox execution (Fly Machines) and workspace storage.
- Amazon Web Services: workspace backups and published pages (S3), and model inference (Bedrock).
- OpenRouter: model inference (see above).
- WorkOS: authentication and directory sync.
- Stripe: subscription billing, payments, and tax compliance.
- Exa: optional web search queries invoked by the agent.
- PostHog: product analytics and error reports, including account identifiers for signed-in users (EU).
- Deepgram: speech-to-text for voice input.
- Composio: connections to third-party apps you choose to connect.
- Steel: the cloud browser the agent uses.
- Resend: transactional and account email.
- Cloudflare: delivery of published pages.
- Apple: push notifications and Sign in with Apple.
- Linear: bug reports you choose to send.
4. Google user data
Pulsar offers an optional integration that connects your Google account so the AI agent can act on the mail, files, and events you already have. This integration is off by default. No Google data is accessed until you complete Google's consent screen and explicitly grant permission.
What we access and why
- Gmail: message content, metadata, and drafts, so the agent can search, summarise, and triage your mail and compose replies you ask it to write.
- Google Drive: the files you direct the agent to, so it can read them into a project and write its output back.
- Google Calendar: events and availability, so the agent can use your schedule as context and create events at your request.
We request the narrowest scopes that support the features you enable. Access tokens are held encrypted at rest and are used only to serve requests you originate.
Limited Use
Pulsar's use and transfer of information received from Google APIs to any other app adheres to the Google API Services User Data Policy, including the Limited Use requirements. Specifically, we do not use Google user data for advertising, we do not sell it, we do not transfer it to third parties except as necessary to provide or improve the feature you are using, to comply with applicable law, or as part of a merger or acquisition, and we do not allow humans to read it unless we have your affirmative consent, it is necessary for security purposes or to comply with applicable law, or the data has been aggregated and anonymised.
Models and Google data
When the agent works on Google content at your request, the relevant portion of that content is sent to the model provider for inference on the same terms described in section 3. Google user data is never used to train, retrain, or fine-tune any AI model, ours or anyone else's.
Retention and revoking access
You can disconnect your Google account at any time from Settings in Pulsar, which revokes our tokens and stops all further access. You can also revoke access from your Google Account permissions page. On disconnection we delete the stored tokens. Google content the agent has already written into your workspace remains there until you delete it, in line with section 5. To request deletion of Google user data we hold, email [email protected].
5. Data retention
Account and billing records are retained for the life of your account plus a limited period required for tax, accounting, or legal purposes. Workspace files persist until you delete them or your account is deleted. Sandbox containers are ephemeral and auto-terminate after idle timeout.
Deleting your account
You can delete your account at any time in Settings → Account, on the web or in the iOS app. This deletes your account, your chats, memories, connections and notification settings, and any workspace only you use, with its projects, files, published pages and apps (cancelling its subscription). In a workspace shared with others, your own chats are deleted and the projects stay with the other members. Billing records are kept, detached from your account, for as long as tax and accounting law requires.
6. Your rights
Depending on your jurisdiction (GDPR, UK GDPR, CCPA/CPRA, India DPDP, etc.) you may have rights to access, correct, export, or delete your personal data, and to object to certain processing. You can exercise these rights by emailing [email protected]. We will respond within 30 days.
7. International transfers
The Service is operated primarily from infrastructure located in the United States. Certain subprocessors process data in the European Union, including product analytics (PostHog). Depending on where you access the Service, your data may be transferred to and processed in the United States or the European Union. Where required (for example, for users in the EEA or UK) we rely on Standard Contractual Clauses and equivalent legal mechanisms.
8. Security
We encrypt data in transit (TLS 1.2+) and at rest. Legacy provider-key credentials, if present on your account, are encrypted with AES-256-GCM using keys stored separately from the ciphertext. No security system is perfect; you use the Service at your own risk.
9. Children
The Service is not directed at children under 16. We do not knowingly collect personal data from children.
10. Changes
We may update this policy from time to time. Material changes will be notified by email or in-app. Continued use of the Service after a change constitutes acceptance.
11. Contact
Starlight Search Inc. · [email protected]
