Skip to content

Privacy Policy

Last updated: 2026-09-11

This Privacy Policy explains how RTC Agent (“the Service”) collects, uses, stores, and protects your personal information. The Service is embedded as a Web Component in third-party websites (such as Mermaid Live Editor) to provide AI assistant capabilities.

When you use the Service, you must authenticate via OAuth2. We collect:

  • OAuth2 user identifier: From your identity provider (e.g., GitHub, Google)
  • Device identifiers: For managing logged-in devices
  • Access and refresh tokens: For maintaining authentication state
  • Message records: Your conversation content with the AI assistant
  • Session metadata: Session title, creation time, status
  • Turn records: Records of each interaction turn
  • RTC call records: Tool call requests and results
  • Session Memory: Context summary of the current session
  • User Memory: Long-term cross-session memory (may include vector embeddings)

The following data is stored only in your browser’s IndexedDB and is never uploaded to our servers:

  • Virtual file system: All file contents (function definitions, scenarios, scripts, etc.)
  • Script execution: All scripts execute locally in the browser sandbox
  • Local sync mirrors: Local copies of sessions/messages/turns (for offline support)

We use collected data for the following purposes:

PurposeData ScopeStorage Location
Provide AI conversation serviceMessage recordsServer
Maintain session contextSession memory, Turn recordsServer
Long-term memory & personalizationUser memoryServer
Authentication & device managementOAuth2 info, Device IDsServer
Local tool executionFiles, Functions, ScriptsBrowser only

We do not upload your local file data to the server. AI tool calls execute in the browser; only execution results (not raw files) are sent back to the server to continue AI reasoning.

  • All file data is stored in your browser via IndexedDB
  • Databases are prefixed with rtc-agent- for user isolation
  • You can clear this data at any time through your browser settings
  • Conversation records, memory, and other data are stored in a PostgreSQL database
  • Transport layer uses WSS/TLS encryption
  • Access tokens are in JWT format; refresh tokens are stored securely

Scripts execute in the browser through a sandboxed engine that:

  • Blocks access to dangerous APIs: fetch, window, document, localStorage, etc.
  • Blocks prototype chain escape (__proto__, constructor)
  • Blocks dynamic import()
  • Blocks unbounded loops
  • Allows only pure computation standard library + rtcAgent host APIs

The Service relies on the following third-party services:

ServicePurposeData Flow
LLM APIAI reasoningConversation content sent to AI model provider
OAuth2 providersAuthenticationAuth info flows to GitHub/Google, etc.
CentrifugeReal-time communicationWebSocket message transport

Please note: When you chat with the AI, your message content is sent to the LLM API for reasoning. Avoid entering highly sensitive information (such as passwords or API keys) in conversations.

You have the following rights:

  • Access: View your account information and conversation records
  • Deletion: Delete your account and all associated data
  • Export: Export your personal data
  • Local data control: Clear all locally stored data through browser settings
  • Account data: Retained until you actively delete your account
  • Conversation records: Retained until session closure or deletion
  • Local data: Retained in your browser until you manually clear it

The Service is not directed to individuals under 18. We do not knowingly collect personal information from minors.

We may update this Privacy Policy from time to time. Updated policies will be posted on this page with the last updated date revised.

If you have questions about this Privacy Policy, please contact us: