Back to Home

AI Transparency & Privacy Center

Last updated: June 2026 · 10 min read · By Harsha Parisha

Transparency Philosophy

Modern developers require absolute clarity when running browser extensions that interact with LLM providers. Because tools like Claude and ChatGPT contain private project repositories, system logs, API credentials, and product strategies, extensions must operate inside sandboxed limits. This Transparency Center lists what Meter AI tracks, what it never touches, its architectural threat model, and the technical standards it obeys.

What Meter AI Collects

  • Interaction Metrics: The count of messages submitted in the active session thread.
  • Context Character Volumes: Character counts inside input prompts and replies to estimate token windows.
  • Reset Timer State: Start times of conversational lockout sequences to predict reset triggers.
  • Local Settings: User preferences, threshold alert configurations, and local sync indicators.
  • License Verification: Subscription level checks processed via secure authentication tokens.

What We Never Collect

  • Prompt Text Content: The raw text strings of your files, codebase imports, or replies.
  • Personal Profiles: Usernames, real names, billing info, or local directory paths.
  • Browser History: Webpages visited outside the explicitly supported LLM interfaces.
  • Remote Telemetries: We do not deploy remote tracking scripts (e.g. Mixpanel, GA) inside the extension.
  • External Auth Tokens: Your provider API keys or session cookies.

Architecture & Data Pipeline

Meter AI is designed as a local-first application. The flow diagram below traces how user input events are captured, processed, and stored entirely in your local browser sandbox, showing that no raw chat transcripts or credentials escape to the cloud.

Meter AI Privacy-First Local Architecture 1. Browser / LLM Tab Claude.ai Web DOM 2. Content Script Local DOM Observers (Chars/Turns) 3. Local Storage chrome.storage.local (Isolated) 4. Popup Panel Extension UI (Renders Metrics) 5. Background Worker Service Worker (Event Hub) 6. License Validation Secure HTTPS Check (Meter AI) DOM Mutation Write Metrics Read Storage Data Flow Runtime Message Optional Check Local Extension Sandbox Boundary (Zero Outbound Leakage)

Independent Standards We Obey

Meter AI's architecture aligns strictly with official browser specifications and global compliance frameworks. We build using the following objective standards to guarantee security:

Permissions Explained

When installing Meter AI, your browser asks for specific permissions. Below is a line-by-line justification for every permission declared in our manifest:

Declared Permission Required Scope Technical Purpose & Safety Scope
host_permissions: ["*://*.claude.ai/*"] Restricted Enables content scripts to run observers on Claude's tab. This is restricted exclusively to the target interface.
activeTab Temporary Allows the extension to read layout geometry and inject DOM styling on demand, avoiding deep background monitoring.
storage Local Only Required to write countdown limits and session history in chrome.storage.local.
declarativeContent Declarative Activates the popup icon automatically when visiting target domains, reducing active background scripts.

Known Limitations & Roadmap

As part of our commitment to transparency, we publish technical constraints and our product roadmap:

Known Limitations

Technical Roadmap

See Also

Trust & Compliance Center Product Specifications Matrix Under the Hood Architecture Guide Learning Center — Understanding Conversational Lockouts