ChatGPT Table of Contents

ChatGPT Table of Contents adds a dependable outline to huge ChatGPT conversations, not just the headings that are visible on screen right now. It builds a prompt-grouped sidebar you can scan, collapse, and use for quick navigation. Sections stay grouped under the prompt that produced them, so long chats remain readable even when the answer spans many parts. When needed, it also uses the current conversation payload from ChatGPT itself so the sidebar can include headings that are not mounted in the viewport yet. That means the outline can stay useful in long threads where ordinary DOM-only TOCs only see the currently loaded window. What you can do with it: - Open a compact TOC sidebar from the floating TOC button. - Jump to the part of the answer you need, even when the target section is buried far down the thread. - Browse headings grouped by prompt instead of losing context in one long page. - Collapse prompt groups and nested headings when you want a shorter outline. - Bookmark headings as Chrome-native bookmarks with a hover-revealed white star that turns black when saved, and copy a lightweight heading reference when you need to cite a section. - Keep your outline stable while switching between ChatGPT conversations in the same tab. - Toggle the sidebar with Cmd+C when you are not editing text or copying a selection. How it works: - Runs only on chatgpt.com and chat.openai.com. - Uses a content script, injected stylesheet, and page-world history bridge inside the active ChatGPT page. - Reads rendered headings, reuses ChatGPT's own authenticated conversation payload for the current chat when available, builds the outline tree, measures layout, and scrolls locally in your browser session. - Uses a minimal event-driven background service worker only to read, create, or remove Chrome-native bookmarks for heading references. - Stores bookmarked heading references as Chrome bookmarks whose URLs include the target heading id, so they can be revisited from Chrome's bookmark manager. - Does not require a separate backend service, popup workflow, analytics, or unrelated third-party API. - Does not send your conversation content to our own servers or unrelated third-party services for outline generation. Who it is for: - Researchers reviewing long multi-step chats. - Engineers and product builders navigating specs, implementation plans, and architecture reviews inside ChatGPT. - Writers organizing structured drafts inside ChatGPT. - Anyone who wants faster navigation in lengthy answers. The extension stays intentionally small: one floating toggle, one sidebar, and one job to do well. ## Changelog ### 1.1.9 - Improved long-thread locate recovery so far-away TOC targets can keep seeking after virtualized content expands instead of stopping on stale geometry. - Tightened unloaded-heading rebound logic so the runtime can recover the intended section from stable heading metadata before weaker text-based fallback kicks in. - Reduced false locate failures near the top of the thread by accepting already-visible synthetic targets and hydrating top-edge matches more reliably. - Fixed long-thread TOC jumps that could fail even when the runtime had already planned a valid target scroll step. - Added host-side runtime transport timeout and reconnect recovery so a stuck runtime port no longer poisons later TOC interactions. - Improved unloaded-heading locate stability by rebinding detached targets from stable heading metadata before falling back to text or ordinal matching. ### 1.1.6 - Added a companion control next to the floating TOC pill so you can collapse or expand all prompt groups at once.