Skip to main content
Privacy is a core feature of Trace. We use a Hybrid Architecture to balance local security with cloud intelligence.

​
🟢 1. Local-First Execution

By default, Trace runs entirely on your machine.
  • SDK/MCP Server: Runs as a local Node.js process.
  • Browser Automation: Uses Puppeteer to control a local Chrome instance.
  • Tools: All 96 tools (click, get_local_storage, read_file) serve data directly from your hard drive/browser to your Agent.
This data never touches our cloud.
  • File contents (trace_read_file) go straight to Claude/Cursor.
  • Cookies and Storage (trace_get_cookies) stay local.

​
🔵 2. Cloud Intelligence (Opt-In)

When you use AI Features (like trace analyze or trace_p), Trace sends specific metadata to the Trace Inference Engine (hosted on Azure/OpenAI). What is sent:
  • Context: Console logs, Network error summaries, and a simplified DOM snapshot.
  • Prompt: The question you asked (e.g., “Why is this failing?”).
What is NOT sent:
  • Cookies/Tokens: We automatically redact common auth headers before transmission.
  • Source Code: We do not upload your source files unless you explicitly ask the AI to “analyze this file”.
  • PII: You are responsible for not asking the AI to analyze screens containing PII.

​
🔐 Authentication

  • API Key: Required for Cloud features. Stored locally in your environment variables.
  • Encryption: All cloud traffic is TLS 1.2+ encrypted.

​
🛡️ Enterprise Mode (Coming Soon)

For strict compliance environments, we will offer:
  • Self-Hosted Engine: Run the inference layer in your own VPC.
  • Local-Only Mode: Disable all cloud calls (tools valid, but AI Analysis disabled).