ReadThread

Explain selected webpage or PDF content in context, continue the conversation, and keep the thread locally.

ReadThread helps you understand difficult passages without losing your place. Select text on a normal webpage and use the configurable keyboard shortcut, context menu, toolbar action, or optional automatic-selection button. ReadThread captures a bounded amount of nearby text, places the exact selection at the center of the request, and opens a focused explanation in Chrome's side panel. Selections inside HTML and ARIA tables gain structured Markdown with headers, captions, spans, units, selected rows or cells, and nearby rows. Selecting a chart title, caption, legend, axis label, or nearby explanation can add one focused visual plus extracted DOM, SVG, and accessibility labels. Simple text tables avoid redundant screenshots. Complex tables, charts, and figures preview one visible-only crop without scrolling the page. Chrome's selection tint is removed before capture. For a broader reading task, choose Get page overview. ReadThread extracts the readable content currently loaded in the page DOM, preserves headings and semantic tables, marks truncation, and excludes common navigation and page chrome. It never scrolls or triggers lazy loading to claim content that was not already available. For PDFs already open in Chrome, select text and choose Explain selection with ReadThread from the context menu to capture exact text plus at most one page-context image. The shortcut can instead detect the visible highlight and crop it automatically. For maximum precision, open a remote URL or local file in the bundled PDF reader and select text or draw around a formula, table, diagram, or scanned passage. Original PDFs are not retained. Continue asking questions while the conversation remains attached to its source. ReadThread supports Markdown and mathematical formulas, reports when a compatible model is reasoning before its answer begins, keeps reading threads locally in IndexedDB, and lets you delete one thread or all local data at any time. Bring your own model credentials. ReadThread supports OpenAI-compatible APIs, Anthropic, Gemini, and Ollama. Provider credentials and requests go directly between the extension and the provider you configure; ReadThread operates no proxy service. You can export a thread as Markdown, JSON, JSONL, CSV, or PDF. A fill-and-review handoff can package the necessary source and conversation context into a draft on ChatGPT, Gemini, Claude, DeepSeek, or Qwen. ReadThread never clicks Send and does not handle platform authentication. The optional automatic-selection button requires website access that you explicitly enable. Shortcut-based capture continues to use temporary access to the current tab. Focused webpage visuals can be excluded from provider requests while remaining local, and text-only providers continue with exact and structured text. ReadThread contains no advertising, analytics, account system, or ReadThread-operated cloud conversation storage. Supported in v0.2: top-frame HTTP and HTTPS HTML pages, bounded loaded-page overviews, selection-anchored tables/charts/figures, and selection-first remote or local PDFs in Chrome 142 or newer. Chrome internal pages, Chrome Web Store pages, cross-origin frames, form fields, rich text editors, OCR, and offscreen webpage rendering are intentionally unsupported.