What Stays Local

A factual breakdown. No marketing language, no hand-waving.

Always local

These are stored on your machine and never leave it through ThinkForge:

  • Project files - All documents, notes, chat logs, and captured content
  • Semantic memory database - SQLite database with embeddings, chunks, tags, and metadata
  • Agent definitions - JSON files defining your agents and their configurations
  • Agent execution outputs - Results from agent runs
  • Agent logs - Execution history and debug information
  • Smart Clipboard contents - Persisted clipboard items
  • Power Strip configuration - Position, layout, commands, per-app settings
  • Workspace settings - Per-project preferences and assistant settings
  • Application settings - Themes, hotkeys, editor preferences

Sent to AI providers when you use them

When you execute an AI action, the following is sent to your configured AI provider:

  • The prompt you wrote or the agent constructed
  • Any context you attached, such as documents, memory chunks, or system instructions
  • Model configuration parameters, such as temperature and token limits
This goes directly to the AI provider. ThinkForge does not proxy, log, or store these requests. We never see your prompts or responses.

Sent to ThinkForge servers

  • Account authentication - Email and auth tokens for login/licensing
  • License validation - Periodic check that your subscription is active

ThinkForge does not currently offer cloud sync. Workspace files, notes, documents, tags, and local organization data are designed to stay on your machine.

Chrome extension

The ThinkForge Chrome extension:

  • Operates only on user-triggered commands, not passive page scraping
  • Captures content only when you explicitly click a capture or save action
  • Sends captured content to the desktop app via local bridge, not over the internet
  • Does not transmit browsing history, page content, or form data to ThinkForge servers

What requires internet

  • AI provider API calls using your keys
  • Account login and license validation
  • Fetching image URLs when using Capture Clipboard on a URL

Everything else is designed to work offline.

Related