Context Recap

Multi-LLM personal memory for ChatGPT, Claude & Gemini — auto-capture, summarize, and inject prior context.

Context Recap — Multi-LLM personal memory for ChatGPT, Claude, and Gemini. Do you find yourself typing the same background information over and over to ChatGPT? Re-explaining a project to Claude that you already discussed with Gemini? Switching between AI assistants and losing context every time? Context Recap solves this. It quietly remembers your past conversations across ChatGPT, Claude, and Gemini, and gently surfaces relevant prior context when you start a new chat — without you ever leaving the LLM site or copy-pasting anything manually. ━━━ How it works ━━━ 1. Install once, give consent once. On first install, an onboarding screen explains what the extension captures, what gets sent externally, and what stays local on your device. You give explicit consent before any capture begins. 2. Use ChatGPT, Claude, or Gemini normally. As you chat, Context Recap automatically captures your messages and the AI's responses — only on these three sites. Each conversation is summarized and tagged in the background. 3. Type a new question. When you start typing 5+ characters in a new conversation, a small floating panel appears in the upper-right corner showing related memories from your past chats — across all three LLM platforms. 4. Click "Use this context". Select one or more memory cards and click the button. The selected context is automatically prepended to your input, so when you press send, the AI gets your full background in one shot. ━━━ Key features ━━━ ✅ Multi-LLM unified memory — Your ChatGPT, Claude, and Gemini conversations live in one personal memory pool. Switching AI assistants no longer means starting from scratch. ✅ Automatic capture & summarization — Conversations are captured as you chat. Summaries, tags, and reaction patterns are extracted in the background using the operator's Firebase Gemini API. No API key from you required. ✅ Semantic search — Memory retrieval uses a multilingual embedding model (Korean and English verified). It finds semantically related memories even when you don't use the exact same keywords. ✅ User control everywhere — The Options page lets you adjust the search threshold, view all your memories, see statistics on what was captured, export/import your data, and delete memories individually or all at once. ✅ Beta notice & transparency — A one-time popup explains the v0.1.0 beta state. The Options page shows a real-time counter of every external transmission (when, what, how many) so you can verify the privacy claims yourself. ━━━ Privacy & data ━━━ 🔒 Zero cost to you — No "Bring Your Own Key" requirement. The operator funds the Gemini API costs, with abuse prevention via per-user quotas and a 30-second emergency kill switch. 🔒 100% local storage — Your memories live in IndexedDB on your device. The operator's server never persists your conversation data — operator-side logs auto-delete after 30 days. 🔒 External transmission only for the enrich step — A short message window (your previous question + previous AI response + current question) is sent to the operator's Firebase Gemini API once per new message. Search, ranking, assembly, and injection all happen 100% locally. 🔒 Zero PII collected — No name, email address, phone number, payment information, or location data. The only identifier is an auto-issued anonymous Firebase UID used for quota enforcement (you never see it; the operator cannot identify you from it). 🔒 No ads, no analytics, no trackers — No advertising SDK, no third-party tracking, no telemetry. Chrome Web Store Limited Use compliant. 🔒 No remote code execution — Manifest V3 hard requirement. All extension code is bundled at build time. The only external file fetched at runtime is the sentence-embedding model from HuggingFace (a static .onnx weights file — data, not code). ━━━ Supported sites ━━━ - ChatGPT (chatgpt.com) - Claude (claude.ai) - Gemini (gemini.google.com) The extension does not access any other website. Removing the extension from Chrome auto-deletes all local data. ━━━ Languages ━━━ 🌍 v0.1.0: Korean & English UI (auto-detected from your system locale). 🌍 v0.1.x patches: 8 additional languages (Japanese, Chinese Simplified/Traditional, Spanish, Portuguese, German, French, Vietnamese) planned incrementally. ━━━ v0.1.0 Open Beta ━━━ 🚧 This is an open beta. We prioritize fast user feedback over UI polish. - Data loss is possible. Incomplete UI is possible. Edge cases will surface. - v0.2 roadmap: master password protection, AES-GCM encryption, full export/import, user-controlled cloud sync (iCloud/Dropbox/Google Drive), Safari/Firefox extensions. - Your feedback shapes the v0.2 priorities. Reach us at contextrecap@gmail.com. ━━━ Operator & contact ━━━ 📧 Operator: Eifer company Contact: contextrecap@gmail.com (response within 1 month per GDPR Article 12(3); within 10 days for Korean users per PIPA Enforcement Decree §41(4)/§43(2)) 📄 Privacy Policy: https://contextrecap.notion.site/Privacy-Policy-356114d05e5d80628fd6dea954570307