# Bulk Delete Memories Source: https://help.memoryplugin.com/api-reference/endpoint/bulk-delete-memories POST /api/memories/bulk-delete Delete multiple memories in a single request. ## Overview Delete multiple memories in a single request. More efficient than calling the single-delete endpoint repeatedly. ### Example ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/memories/bulk-delete \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "memoryIds": ["mem-id-1", "mem-id-2", "mem-id-3"] }' ``` # Create Memory Source: https://help.memoryplugin.com/api-reference/endpoint/create POST /api/memory # Create Bucket Source: https://help.memoryplugin.com/api-reference/endpoint/create-bucket POST /api/buckets # Delete Conversations Source: https://help.memoryplugin.com/api-reference/endpoint/delete-conversations DELETE /api/chat-history/chats Delete one or more conversations and their associated data (messages, chunks, vectors). ## Overview Delete one or more conversations and all associated data (messages, chunks, and vector embeddings). This action is irreversible. ### Example ```bash theme={null} curl -X DELETE https://www.memoryplugin.com/api/chat-history/chats \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "ids": ["conv-id-1", "conv-id-2"] }' ``` # Delete Memory Source: https://help.memoryplugin.com/api-reference/endpoint/delete-memory DELETE /api/memories/{memoryId} Delete a single memory by its ID. Removes the memory from the database and vector store. ## Overview Delete a single memory by its ID. The memory is removed from both the database and the vector store. ### Example ```bash theme={null} curl -X DELETE https://www.memoryplugin.com/api/memories/mem-abc-123 \ -H "Authorization: Bearer YOUR_API_KEY" ``` # Get Memories Source: https://help.memoryplugin.com/api-reference/endpoint/get GET /api/memory ## Image Memories Responses may include image memories alongside text memories. Image memories include additional fields: * `content_type: "image"`: identifies this as an image memory * `image_url`: a signed URL to view the image (expires after 4 hours) * `image_description`: AI-generated description of the image content ### Example Response ```json theme={null} { "data": [ { "id": "user-id-uuid", "text": "User prefers dark mode in all applications", "content_type": "text", "metadata": { "text": "User prefers dark mode in all applications", "createdAt": "2026-04-01T10:00:00Z", "bucketId": 65, "bucketName": "General" } }, { "id": "user-id-uuid-2", "text": "A whiteboard diagram showing the auth flow with three microservices.", "content_type": "image", "image_url": "https://memoryplugin.com/storage/v1/object/sign/memory-images/...", "image_description": "A whiteboard diagram showing the auth flow with three microservices.", "metadata": { "text": "A whiteboard diagram showing the auth flow with three microservices.", "createdAt": "2026-04-06T10:30:00Z", "bucketId": 65, "bucketName": "General" } } ], "count": 2 } ``` Image URLs are temporary signed URLs that expire after 4 hours. If you need to access an image after the URL has expired, make another API request to get a fresh URL. # Get Buckets Source: https://help.memoryplugin.com/api-reference/endpoint/get-buckets GET /api/buckets # Get Conversation Source: https://help.memoryplugin.com/api-reference/endpoint/get-conversation GET /api/chat-history/conversation Retrieve a specific conversation with all its messages. ## Overview Retrieve a specific conversation with all its messages. Use the `conversationId` from the list or search endpoints. ### Example ```bash theme={null} curl "https://www.memoryplugin.com/api/chat-history/conversation?conversationId=abc-123" \ -H "Authorization: Bearer YOUR_API_KEY" ``` # Get Conversation Count Source: https://help.memoryplugin.com/api-reference/endpoint/get-conversation-count GET /api/chat-history/count Returns the total count of imported conversations (excluding excluded and quota-blocked ones). ## Overview Returns the total count of imported conversations for the authenticated user. Excludes conversations that have been excluded or blocked by quota limits. ### Example ```bash theme={null} curl "https://www.memoryplugin.com/api/chat-history/count" \ -H "Authorization: Bearer YOUR_API_KEY" ``` # List Conversations Source: https://help.memoryplugin.com/api-reference/endpoint/list-conversations GET /api/chat-history/chats List imported chat conversations with pagination and filtering. ## Overview List imported chat conversations with pagination and filtering. Use this to browse conversations, filter by platform, or search by title. ### Example ```bash theme={null} curl "https://www.memoryplugin.com/api/chat-history/chats?page=1&pageSize=10&provider=chatgpt" \ -H "Authorization: Bearer YOUR_API_KEY" ``` # Recall Chat History Source: https://help.memoryplugin.com/api-reference/endpoint/recall-chat-history POST /api/chat-history/inject Search and synthesize context from the user's past AI conversations. This is the primary recall endpoint — it performs semantic search, then uses AI to synthesize matching results into a coherent summary with source citations. This is the same function powering the `recall_chat_history` tool in the MCP server. Supports parallel queries for complex topics that benefit from multiple search angles (e.g., timeline, decisions, people). ## Overview This is the primary endpoint for retrieving context from a user's past AI conversations. It searches across all imported chat history, then uses AI to synthesize the results into a coherent summary with source citations. This is the same function that powers the `recall_chat_history` tool in the [MCP server](/integrations/mcp-server). ### How it works 1. Your query is embedded and searched against all conversation chunks (vector + keyword hybrid search) 2. Results are filtered and ranked by relevance 3. An AI model synthesizes the top results into a coherent narrative 4. Source citations are attached so you can trace back to specific conversations ### When to use this vs. Search | Endpoint | Returns | Best for | | - | - | - | | **Recall** (`/inject`) | AI-synthesized summary + sources | Building context for AI conversations, getting a coherent answer | | **Search** (`/search`) | Raw matched chunks with scores | Building custom UIs, debugging, fine-grained control | ### Parallel queries For complex topics, use the `queries` array to search from multiple angles simultaneously. Each query is searched and synthesized independently. ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/chat-history/inject \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "queries": [ { "query": "what database schema decisions were made", "maxTokens": 500 }, { "query": "authentication architecture discussions", "maxTokens": 500 } ] }' ``` ### Single query ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/chat-history/inject \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "query": "what did I decide about the API rate limiting approach?", "maxTokens": 800 }' ``` # Search Chat History Source: https://help.memoryplugin.com/api-reference/endpoint/search-chat-history POST /api/chat-history/search Perform raw semantic search across all imported chat conversations. Returns matched chunks with scores and surrounding context. For most use cases, use the Recall endpoint (`POST /api/chat-history/inject`) instead — it performs search and AI synthesis in one step. ## Overview Perform semantic search across all imported chat conversations. This endpoint uses hybrid vector + keyword search with AI reranking for high relevance results. ### How it works 1. Your query is embedded and compared against all conversation chunks using vector similarity 2. Results are enriched with surrounding context messages 3. An AI reranker scores and orders results by relevance (can be skipped for faster responses) ### Example ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/chat-history/search \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "query": "how to set up authentication", "limit": 10, "provider": "all" }' ``` # Search Memories V2 Source: https://help.memoryplugin.com/api-reference/endpoint/search-memories GET /api/v2/memory ## Image Memories Search results may include image memories alongside text memories. Image memories are returned as objects with additional fields: * `content_type: "image"`: identifies this as an image memory * `image_url`: a signed URL to view the image (expires after 4 hours) * `image_description`: AI-generated description of the image content Image memories are ranked by relevance alongside text memories using multimodal embeddings that enable cross-modal search (e.g., a text query can find visually relevant images). ### Filtering by Content Type Use the `contentType` query parameter to filter results: ```bash theme={null} # Only text memories curl "https://www.memoryplugin.com/api/v2/memory?query=architecture&contentType=text" \ -H "Authorization: Bearer YOUR_API_KEY" # Only image memories curl "https://www.memoryplugin.com/api/v2/memory?query=diagram&contentType=image" \ -H "Authorization: Bearer YOUR_API_KEY" # All memories (default) curl "https://www.memoryplugin.com/api/v2/memory?query=architecture" \ -H "Authorization: Bearer YOUR_API_KEY" ``` ### Example Response with Image Memories ```json theme={null} { "memories": [ "User prefers dark mode in all applications", { "content_type": "image", "image_url": "https://memoryplugin.com/storage/v1/object/sign/memory-images/...", "image_description": "A whiteboard diagram showing the auth flow with three microservices connected via Redis pub/sub.", "metadata": { "createdAt": "2026-04-06T10:30:00Z", "bucketId": 65, "bucketName": "Work" } }, "Meeting notes: decided to use PostgreSQL for the new service" ], "buckets": [ { "id": 65, "name": "Work" }, { "id": 66, "name": "Personal" } ] } ``` Image URLs are temporary signed URLs that expire after 4 hours. If you need to access an image after the URL has expired, make another API request to get a fresh URL. # Update or Move Memory Source: https://help.memoryplugin.com/api-reference/endpoint/update-memory POST /api/v2/memory/update ## Overview Update a memory's text, move a memory to a different bucket, or do both in a single request. You can also move up to 100 memories into one bucket at once with the bulk form. This is the REST equivalent of the `update_or_move_memories` MCP tool. Both share the same behaviour. ### Authentication Send your API key as a bearer token: ``` Authorization: Bearer YOUR_API_KEY ``` ## Single memory update Provide a `memoryId` plus at least one of `text`, `bucketId`, or `bucketName`. If you send only a `memoryId` with none of these fields, the request is rejected. ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/v2/memory/update \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "memoryId": "mem-abc-123", "text": "User moved to San Francisco and now works at Stripe", "bucketName": "Work" }' ``` The response returns the updated memory: ```json theme={null} { "success": true, "memory": { "id": "mem-abc-123", "text": "User moved to San Francisco and now works at Stripe", "bucketId": 65, "version": 3, "updatedAt": "2026-07-04T10:30:00Z", "createdAt": "2026-04-06T10:30:00Z" } } ``` Changing a memory's text re-embeds it, so search stays accurate against the new wording. Each text or bucket change increments the memory's `version` and records a snapshot in its edit history, so you can see what changed and when. ## Bulk move To move several memories into the same bucket, send a `memoryIds` array (1 to 100 IDs) with either `bucketId` or `bucketName`. Bulk requests move memories only; they do not change text. ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/v2/memory/update \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "memoryIds": ["mem-id-1", "mem-id-2", "mem-id-3"], "bucketName": "Work" }' ``` ```json theme={null} { "success": true, "updated": 3, "bucketId": 65 } ``` Bulk moves are all-or-nothing. If any ID in the list cannot be found, the whole request is rejected with a `404` and the unresolved IDs are returned so you can retry: ```json theme={null} { "error": "Some memory IDs could not be resolved", "rejectedIds": ["mem-id-2"] } ``` ## Targeting a bucket You can point at a destination bucket two ways: * **`bucketId`**: a numeric bucket ID. The bucket must already exist, and you need write access to it. * **`bucketName`**: a bucket name. If no bucket with that name exists, it is created for you and the memory is moved into it. Send one or the other, not both. Provide either `memoryId` (single) or `memoryIds` (bulk) in a request, never both. # Upload Chat History Source: https://help.memoryplugin.com/api-reference/endpoint/upload-chat-history POST /api/chat-history/ingest/custom-online Upload a single conversation from any platform. Supports upsert semantics — re-sending a conversation with the same ID will update it with new messages. The conversation is processed asynchronously (chunked, embedded, and indexed for search). ## Overview Upload chat history from any platform to MemoryPlugin. This endpoint accepts a single conversation per request with your own unique identifiers for conversations and messages. ### Key features * **Upsert semantics**: Re-sending a conversation with the same `conversation.id` updates it with any new messages rather than duplicating it. * **Dynamic platform naming**: Supply a `platform` identifier to group conversations by source. Each platform is scoped to your user account. * **Full processing pipeline**: Conversations are chunked, embedded, and indexed for semantic search, just like conversations imported via the browser extension. ### Limits * Conversations exceeding **100,000 tokens** are rejected during background processing. * Maximum payload size: **50 MB**. * Platform identifier: alphanumeric, hyphens, underscores, dots. Max 100 characters. ### Processing Conversations are processed **asynchronously**. The endpoint returns immediately with `status: "queued"`. Background processing (chunking, embedding, vector indexing) typically completes within a few seconds for normal-sized conversations. ### Example ```bash theme={null} curl -X POST https://www.memoryplugin.com/api/chat-history/ingest/custom-online \ -H "Authorization: Bearer YOUR_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "platform": "my-slack-bot", "platformDisplayName": "My Slack Bot", "conversation": { "id": "conv-abc-123", "title": "Discussion about architecture", "createdAt": "2026-04-27T10:00:00Z", "updatedAt": "2026-04-27T11:30:00Z", "messages": [ { "id": "msg-1", "role": "human", "content": "How should we structure the API?", "createdAt": "2026-04-27T10:00:00Z" }, { "id": "msg-2", "role": "assistant", "content": "I would recommend a RESTful approach with versioned endpoints...", "createdAt": "2026-04-27T10:01:00Z" } ] } }' ``` # API Introduction Source: https://help.memoryplugin.com/api-reference/introduction Learn how to integrate with the MemoryPlugin API ## Welcome to the MemoryPlugin API The MemoryPlugin API lets you store, retrieve, and search memories programmatically, and query your imported chat history. It is the same API the browser extension and integrations use. View the complete OpenAPI specification ## Base URL All API requests should be made to: ``` https://www.memoryplugin.com ``` ## Authentication Every endpoint requires your API token. Get it from the [dashboard](https://memoryplugin.com/dashboard/connect?tab=access): click **Connect** in the sidebar, open the **Access** tab, and copy your token. Include it as a bearer token in the request headers: ```bash theme={null} Authorization: Bearer YOUR_API_TOKEN ``` Treat the token like a password. It grants full access to your memories and chat history, and you can regenerate it from the same Access tab if it leaks. ## Response Format Endpoints return plain JSON objects. There is no shared envelope; each endpoint documents its own response shape. For example, `GET /api/v2/memory` returns: ```json theme={null} { "memories": [ "Prefers concise answers without preamble", "Works as a product designer at a fintech startup" ], "buckets": [ { "id": 1, "name": "General" }, { "id": 2, "name": "Work" } ] } ``` ## Error Handling Errors use standard HTTP status codes (`400` for invalid input, `401` for a missing or bad token, `404` when something does not exist) with a JSON body: ```json theme={null} { "error": "Human-readable error message" } ``` ## Available Endpoints * **Memory endpoints**: store, retrieve, search, update, and delete memories * **Bucket endpoints**: list and create memory buckets * **Chat history endpoints**: recall, search, upload, and manage imported conversations Explore the endpoint documentation for request and response details. # Changelog Source: https://help.memoryplugin.com/changelog Product updates and announcements for MemoryPlugin ### Chat History, one section Browse, Search, Imports, and Activity used to feel like separate pages. They're now one Chat History section with a shared header, so you can jump between them without losing your place. Filters sit as chips along the top. Search covers every tool you sync, not just a handful, and sorting by most recent actually works. On Imports, drag an export onto the page to start an upload. See the [chat history dashboard](/features/chat-history/dashboard). ### ChatGPT Project chats now sync (Extension v4.2.3) Conversations you keep inside ChatGPT Projects were invisible to sync. They show up now, same as the rest of your history. Auto-sync, Chat Sync, and Sync Chats all pick them up. ### Switching ChatGPT accounts no longer mixes your chats (Extension v4.2.4) Jump between ChatGPT accounts or workspaces and the extension used to keep syncing the old one. It now follows you within a few seconds. If you have a lot of Project chats, sync also backs off when ChatGPT asks it to slow down, instead of hammering and getting rate-limited. ### MemoryPlugin Sync 2.2.0 The Mac app was signing you out when the main window, quick search, and a second copy all refreshed at once. They now share one session. Headless Claude Code automation runs no longer bury your real conversations under thousands of script sessions. Switching accounts from another window is clean. Needs macOS 12.3 or later. See the [desktop app docs](/integrations/desktop-app). ### A native iPhone app The iOS app used to be a thin shell around the Safari extension. It's now a real app: * **Ask, on your phone** - question your memories and chat history from anywhere. Answers start appearing right away, with a quick note on what it's searching while it works. * **Search everything** - one search box over your entire chat history, with each tool's real logo on its conversations so you can tell them apart at a glance. * **A reader built for scrubbing** - open any conversation and the timeline shows the whole chat at once; drag along it to preview each turn and jump straight where you want. Timestamps and day separators keep long chats legible. * **Your memories, natively** - browse buckets, add, edit, move, and delete memories, review Smart Memory categories, and work through cleanup Suggestions. * **Pin from anywhere** - pin conversations right from search results or the reader. The Safari extension is still included for memory on AI sites in your mobile browser. The app is in beta on [TestFlight](https://testflight.apple.com/join/3nBR4xdN). ### Claude's chunked exports now import cleanly Claude changed its data export format: larger accounts get a manifest file plus several ZIP archives instead of one JSON. The importer now understands all of it. Select every ZIP at once and upload them together; the import modal walks you through where the files come from. And if you upload the manifest itself by mistake, the error tells you exactly what it is and what to upload instead. ### Imports email you now Large imports can take hours, so you no longer have to babysit them: you get one email when processing starts and one when it finishes, with any problem files listed. Start the import, close the tab, come back when it's done. ### ChatGPT memory capture fixed (Extension v4.2.1 and v4.2.2) In early August ChatGPT redesigned its conversation page, which quietly stopped the extension from saving new memories there. v4.2.1 fixes it. v4.2.2 makes the whole category of breakage less painful: when any supported site redesigns, we can now ship the repair server-side, so it reaches you in about an hour instead of waiting days for an extension store review. ### Smoother MCP connections Three fixes for connecting AI tools over MCP: newly connected tools can save memories right away instead of hitting a permissions error, some clients that failed during sign-in because of an overly strict handshake check now connect fine, and tools show clean display names in Claude, Cursor, and other clients instead of function names. ### Claude Memory Visualizer A new free tool that turns Claude's exported memory file into something you can actually read. Try it at [memoryplugin.com/tools/claude-memory-export](https://www.memoryplugin.com/tools/claude-memory-export). ### Fixes * **Current chat in Chat Sync** - updates when you switch conversations, instead of showing the one you just left (Extension v4.2.5). * **Sidebar sync bar** - the extra sync bar we injected into the ChatGPT and Claude sidebars is hidden for now. The per-chat dots are still there, and sync still runs from the panel. * Connecting MemoryPlugin in Cursor or VS Code no longer fails during sign-in. * The chat history import dialog now stays fully on screen on short laptop displays. * Selecting text on dark dashboard surfaces is visible again. ### MemoryPlugin Sync: your coding chats, in your memory Claude Code, Codex, and Cursor conversations live in files on your Mac, invisible to every other AI tool you use. **MemoryPlugin Sync**, our new macOS desktop app, changes that. Sign in, tick the projects and chats you want, and it keeps them synced into your chat history in the background from then on. Privacy is the point: only your prompts and the assistant's replies sync. Tool output, file contents, terminal commands, and thinking never leave your machine. The app moved fast after its July 8 launch, and by month's end it had grown into much more than a sync tool: * **The full dashboard, embedded** - memories, Ask, chat history, and imports in the app, no browser needed. * **Global search with ⌘⇧K** - a search window that opens over any app on your Mac and covers your entire synced history, ChatGPT and Claude chats included, not just coding tools. Filter by tool, jump to the chat in MemoryPlugin or on the platform it came from. * **Projects stay attached** - every synced coding chat carries its repo or project name, and the dashboard's chat history gets a Project filter to match. * **Focused sync** - by default the app syncs your prompts and each turn's final answer; the agent's play-by-play stays local. Turn it off if you want everything. * **Full per-chat control** - pause auto-sync from the menu bar, exclude any chat, or **Remove from MemoryPlugin** to delete it from your history and stop it syncing again (the local file stays put). * **Signed, notarized, and self-updating** - the app checks for updates daily and installs them only when you say so. [Download for macOS](https://www.memoryplugin.com/download) (Apple Silicon, macOS 12 or later; Intel and Windows are on the way). Syncing requires Pro; browsing your local chats in the app is free. Read the announcement: [Meet MemoryPlugin Sync](https://blog.memoryplugin.com/meet-memoryplugin-sync/), or see the [desktop app docs](/integrations/desktop-app). ### Quality mode recall, rebuilt When an AI tool recalls from your chat history over MCP, it can ask for `speed` or `quality` mode. Quality mode was rebuilt end to end this month. Answers are now grounded in what your conversations actually said (in our evals, the hallucination rate dropped from 48% to 30%), read as connected prose instead of clipped bullet fragments, and dig through roughly twice as many candidate conversations without silently dropping the most relevant one. A reworked pipeline made all of that faster too: median latency fell from 6.9s to 4.6s. The honest catch remains: quality mode is still the slower option, by design. Quality mode is a Pro feature; on Core, recalls run in speed mode. More on the thinking behind it: [Quality Mode and context engineering](https://blog.memoryplugin.com/context-engineering-claude-5-quality-mode/). ### A real reader for your chat history Opening an archived conversation used to mean scrolling a wall of chat bubbles. It's now a proper transcript: a clean reading column, each turn under a speaker line that stays visible as you scroll, and full-contrast text for both sides of the conversation. Longer conversations (8+ turns) also get a **conversation map** on wide screens: a slim rail along the right edge with one mark per message, sized by length and colored by speaker. Touch it and a list of every turn opens beside it; click anywhere to jump. Search matches light up on the map too. And any conversation can now leave with you: **Export as Markdown** or **Export as JSON** from the reader's menu. ### Speak your memories There's a microphone button next to the Add Memory box on the dashboard now. Talk for up to five minutes; MemoryPlugin transcribes the recording and turns it into clean, individual memories that you review and edit before anything is saved. Available on all plans. ### The dashboard, reorganized The sidebar was rebuilt around what you actually do: your two libraries (memories and chats) up top, with everything else in its place. A new **Connect** section shows every AI tool linked to your account with live status, so you can see at a glance where MemoryPlugin is active, and a first-run setup wizard gets new accounts from signup to a working connection. The Imports page was redesigned to match, and your API token now lives under **Connect → Access**. ### Easier imports * **Upload export ZIPs as-is** - Claude and ChatGPT export archives no longer need unzipping. ZIPs up to 512 MB, JSON up to 1 GB. * **Errors that name the fix** - upload the wrong file from an export (like `memories.json`) and the message tells you what that file is and which one to upload instead. * **Paste text as memories** - a new tab in the Import Memories dialog: paste notes or markdown, and an AI pass splits them into individual memories you review before saving. * **Friendlier import files** - memory imports accept a simple `memory` column, plain one-per-line text files, and come with downloadable CSV and JSON samples. * **A redesigned Import Memories dialog** that confirms which bucket you're saving to before anything is written. ### A hub of free tools A new hub at [memoryplugin.com/tools](https://www.memoryplugin.com/tools) collects free, no-signup utilities that run entirely in your browser: converters that turn ChatGPT, Claude, Grok, and TypingMind exports into clean Markdown (plus PDF and Obsidian flavors for ChatGPT), an AI chat export viewer, a Claude Code session viewer, a ChatGPT memory exporter, an AI code extractor, and **AI Wrapped**, a year-in-review for your AI usage. Nothing you load into them is uploaded anywhere. ### The Trust page "Can you read my chats?" is a fair question to ask any memory product, so we answered it in public. [memoryplugin.com/trust](https://www.memoryplugin.com/trust) covers what we store, what we never see, and what leaves your machine, in plain language rather than legal boilerplate. ### MCP server: open source and in the registry * The MCP server is now **open source** at [github.com/memoryplugin/mcp-server](https://github.com/memoryplugin/mcp-server) and published on npm as `@memoryplugin/mcp-server`. * MemoryPlugin is listed in the **official MCP Registry** as `com.memoryplugin/memory`, so registry-aware clients can find it directly. * New tool: **`export_conversation`** - your AI can hand you a complete conversation as a downloadable JSON file via a short-lived link. * A public reference page is live at [memoryplugin.com/mcp](https://www.memoryplugin.com/mcp). ### Agent skills for coding agents One open-source install teaches your coding agents to recall what they need before assuming, and store what matters before finishing. Claude Code installs it as a plugin in one command; `npx skills add memoryplugin/agent-skills` covers Codex, Cursor, GitHub Copilot, Gemini CLI, and other skills-capable agents. MIT licensed, on GitHub at [memoryplugin/agent-skills](https://github.com/memoryplugin/agent-skills). Details at [memoryplugin.com/agent-skill](https://www.memoryplugin.com/agent-skill). ### Account and billing * **Self-service refunds** - if a subscription charge surprises you, request a refund yourself from your account settings within 14 days of the charge. The subscription cancels immediately and the refund follows, no support ticket needed. * **Reminder emails before charges** - a heads-up before a trial converts and before renewals, so a charge is never a surprise. * **OTP sign-in** - resending a login code now works reliably. ### Extension v4.1.0 and v4.1.1 The extension now tells the dashboard which sites it's active on, which is what powers the Connect board's live status. Also: leaner analytics and an uninstall feedback page. ### Fixes * **Gemini conversations keep all their turns** - synced Gemini chats were collapsing to a single exchange; they now keep the full conversation in order, and existing imports were repaired. * **Summaries unblocked** - some conversations were wrongly flagged as too large to summarize; that's fixed, including for chats already in your history. * **Fewer transient errors** - momentary database hiccups in the dashboard now retry automatically instead of surfacing as errors. * **Activity heatmap** no longer times out on accounts with years of history. ### Shared memory buckets You can now share a memory bucket with someone else. Invite them by email as a Viewer, Contributor, or Editor, and every memory in the bucket shows who added it. Handy for a partner, an assistant, or a small team that wants their AIs working from the same notes. Sharing is available on the Pro plan; see the [docs](/features/shared-buckets). ### Search your chat history from any tab (Extension v4.0.0) A full-screen chat history search is now built into the extension. Press **⌘+Shift+K** (**Ctrl+Shift+K** on Windows and Linux) from any supported site to search everything you've synced, browse recent conversations, filter by tool, star the chats you want to keep close, and read the full transcript of any conversation. You can also open the search from the History button in the panel. ### Two new platforms Memory injection, smart recall, and search now work on **NotebookLM** and **Xiaomi MiMo AI Studio**, bringing MemoryPlugin to two more places you already chat with AI. ### Extension v4.0.0 The recent run of extension updates - the panel redesign with tabs, AI-driven recall, and the Extract memory selection pill - is now rolled into a single stable **4.0.0** release. If you had an earlier version, you'll be brought up to date automatically. ### The Field Guide to AI Memory A new, free guide to how AI memory actually works is live at [memoryplugin.com/wiki](https://memoryplugin.com/wiki). It covers the techniques, the tradeoffs, and honest numbers, written for anyone who wants to understand what's happening under the hood. ### Comparison pages New pages walk through how MemoryPlugin compares to other approaches to AI memory, alongside a refresh of the platform pages that show what works where. ### Edit memories from anywhere A new memory editing tool is available across MCP and the REST API. Connected AI tools can now update or move existing memories — fixing details, rewording, or shifting a memory to a different bucket — without you needing to step into the dashboard. * **MCP `update_or_move_memories`** — available in all MCP-connected clients (Claude, ChatGPT, Cursor, and more) * **`POST /api/v2/memory/update`** — for direct API integrations ### Memory version history Every memory now keeps a full history of its edits. Open a memory to see when it changed, what was added or removed (highlighted inline), and who or what made the change — you, an AI tool, or an automated improvement. **⌘+Enter / Ctrl+Enter** saves quickly without leaving the keyboard. ### Selection-to-memory pill (Extension v3.5.0) Highlight any text inside a chat message on ChatGPT, Claude, Gemini, Grok, DeepSeek, Kimi, MiniMax, or TypingMind and an **Extract memory** pill appears next to your selection. One click turns the highlighted text into a clean memory and saves it to your current bucket. ### Multi-summary chat history Generate multiple summaries per conversation with **custom prompts** — get a meeting-notes view, an action-items view, and a key-decisions view of the same chat, each tailored to how you want to read it. Switch between summaries from the chat panel. ### Word document uploads in file buckets File buckets now accept `.docx` and `.doc` files alongside PDF, Markdown, and plain text. Word docs are processed with the same page-aware chunking as PDFs, so citations point to the right place. ### Smarter chat history recall The chat history recall tool now supports: * **Time filters** — ask the AI to recall only what you said in the last week, last month, or any custom date range * **Speed vs. quality mode** — choose faster responses for quick lookups, or higher-quality reasoning for complex questions where accuracy matters more than latency ### AI-driven chat history recall (Extension v3.3.0) Recall is now smarter and quieter. Instead of searching your chat history every time you send a message, the AI decides when context from past conversations would help and asks for it. A preview panel shows what's being searched and lets you preview or cancel the injection before it's sent. The old auto/ask mode selector has been retired — recall now fires only when the AI requests it. ### Extension v3.4.0 — Panel redesign & real-time updates * **Real-time updates** — the panel refreshes automatically when memories are added, edited, or deleted on another device. No more manual refresh. * **Instant login/logout detection** — auth state syncs across tabs immediately. * **Account tab on all browsers** — previously Safari-only. Shows your avatar, plan, and quick links to manage your subscription. * **Avatar in panel header** — your profile picture appears at the top of the panel and opens your account view. * **Contextual upgrade prompts** — free → paid, Core → Pro, Pro → Lifetime, all shown in-extension. * **Cleaner header** — Refresh, Copy, and Load-limit controls moved into a tidy overflow menu on the memories toolbar. * **Faster page loads** — the extension is significantly lighter on every page it runs on. ### Plan badges in the extension The extension header now shows your plan — **Core**, **Pro**, or **Pro LT** (Lifetime) — right next to the MemoryPlugin title. ### Memory update toast (Extension v3.2.0) Return to a chat with new or edited memories since you last visited it and you'll see a toast like *"2 new, 1 updated."* Click **Re-inject memories** and only the changed memories are sent into the chat with a contextual note — no need to flood the conversation with your full memory set. ### Mobile overflow menu (Extension v3.2.0) Copy and Export collapse into a `⋯` menu on narrow viewports so the panel header doesn't crowd on mobile. ### Live dashboard Edit a memory from the extension, MCP, or the API and the dashboard reflects the change instantly — no more polling delay. ### Add memory modal: bulk paste & image uploads The dashboard's Add Memory dialog now accepts multiple memories at once (one per line) and supports drag-and-drop image uploads in the same flow. ### Sidebar bucket sections The dashboard sidebar's bucket list has been unified into a single section with inline sort controls — pick recency or custom order without diving into settings. ### Image search controls * **Filter search to images or text only** — both from the dashboard and via the memory search API * **More relevant image search results** with an improved ranking approach ### Landing page refresh New hero, FAQ section, testimonials, and a dedicated chat history section on [memoryplugin.com](https://memoryplugin.com). ### Fixes * **DeepSeek chat sync on Safari** restored (Extension v3.4.1) * **TypingMind chat history imports** no longer fail on large uploads * **Image memory deletion** now fully removes the image, including from search * **Custom-provider chat history** now displays with its own platform identity instead of being mislabelled * **Subscription changes** no longer accidentally downgrade your plan * **Lifetime plan detection** is now reliable across the web app and extension ### Image memories Memories can now be images. Upload a photo, screenshot, or diagram and MemoryPlugin will: * **Understand what's in it** so you can find it later by describing what you remember * **Search alongside your text memories** in a single unified result list * **Show thumbnails inline** in your memory list, with a full-size lightbox view on click Image memories work everywhere your text memories do — search, MCP, bucket organization, and deletion. Upload from the dashboard's Add Memory dialog or the browser extension. ### Custom platform chat history Third-party integrations and self-hosted platforms can now send conversations into MemoryPlugin through a new generic upload endpoint. Conversations from custom platforms appear throughout the dashboard with their own platform identity, not lumped under an existing provider. ### Pro → Lifetime upgrade in settings Pro subscribers can now upgrade to a Lifetime plan directly from Settings, with the price adjusted for what you've already paid on your existing subscription. ### Extension v3.1.5 — Safari sync reliability Heavy sync users on Safari were running into storage errors that interrupted background sync. The extension now stores its cache without that limit, so syncing stays reliable on Safari. Also fixed: the panel showing "Loading…" instead of memories that were already cached — they now appear instantly. ### Extension v3.1.4 — Cross-tab auth & empty states * **Logout propagates instantly across all tabs** — no more browser reload needed to clear auth state * **Logged-out panel** now shows a sign-in prompt instead of a stale loading spinner * **Empty-state UI** for users with zero memories — shows guidance, an "Add a memory" button, and help links * **Inline Add button** in the memories header for quick saves without leaving the panel ### Reliability * **Self-healing chat imports** — conversations that imported without their messages are now automatically repaired in the background * **Failed imports no longer go quiet** — they're now correctly flagged as failed so you can retry them * **Memory delete API** accepts either the internal ID or the public `pcId`, so any client can delete a memory with whatever identifier it has * **Deleted memories no longer leave orphan entries** in search ### Activity page (beta) New analytics dashboard for your chat history: * **At-a-glance metrics** — total conversations, messages, and active days * **Conversations over time** — see how your activity trends across weeks and months * **Platform breakdown** — visualize your usage across ChatGPT, Claude, and more * **Activity heatmap** — GitHub-style grid showing when you chat most * **Time range selector** — filter by week, month, or all time ### Monthly Insights AI-generated reflections on your Activity page. Each month, MemoryPlugin reviews your conversations and surfaces recurring themes, projects, and interests — only topics that come up across multiple conversations are included, keeping the signal high. ### Grok chat history import Import your Grok conversations with full support for branching threads. Available as both file upload and online sync from the browser extension. ### Gemini chat history import & auto-sync Import and auto-sync your Google Gemini conversations. Sync runs in the background with configurable limits and cancellation support. ### DeepSeek online sync Sync your DeepSeek conversations directly from the browser extension — no export files needed. ### Search improvements * **Better relevance** — upgraded to a larger, more reliable model for ranking results * **Smarter query understanding** — improved detection of what you're looking for * **Keyword matching** — results that match more of your search terms rank higher * **Fewer missed results** — relaxed filtering keeps contextually relevant results that were previously excluded ### Imports page redesign * Platform-colored cards so each import is identifiable at a glance * Clear status badges for processing, completed, and failed imports * Summary stats showing total files, conversations, and messages * Filters to narrow by platform, status, or date range ### Multi-file ChatGPT imports Upload multiple ChatGPT export files at once, with guidance for splitting large exports. ### Browser extension v3.1.1 & v3.1.2 * Renamed "Sync" to "Sync All" for clarity * Fixed login flow not updating the UI after signing in on Chrome ### Per-bucket Knowledge Graph Each bucket now has its own knowledge graph view. MemoryPlugin extracts the key concepts across the memories in a bucket and visualizes how they connect — useful for seeing the shape of what you've collected on a topic. The graph was rebuilt for smoother interaction with larger buckets. ### AI-powered docs help on the dashboard A floating help button on the dashboard opens an in-app chat that answers questions about MemoryPlugin using the documentation as context. Responses include source links and suggested follow-up questions, so you can get answers without leaving the app. ### More reliable Monthly Insights Monthly Insights / Life Context generation now runs as a dedicated background job with status tracking, so monthly summaries arrive more reliably and you can see when they're being prepared. ### Extension onboarding Interactive walkthrough for new users: * **Essentials** — four chapters covering core features with realistic mock UIs * **Progress tracking** — pick up where you left off * Opens automatically on first install; also accessible from Settings ### Sync panel redesign Compact icon-based provider selector with platform logos — in color when active, grayscale when inactive. Scales better as we add more platforms. ### ChatGPT to Markdown converter Free tool for converting ChatGPT export files to clean, readable Markdown. Runs entirely in your browser — nothing is uploaded. Handles even very large export files without freezing. Try it at [memoryplugin.com/tools/chatgpt-to-markdown](https://memoryplugin.com/tools/chatgpt-to-markdown) ### Chat import reliability * Better handling of very large conversations * Retry failed imports directly from the imports page * Real-time progress tracking during sync * Improved error recovery and cancellation ### Browser extension v3.0.8 → v3.1.0 * Grok voice mode support * LibreChat compatibility fix * Bucket ordering now matches your dashboard preferences * General stability improvements ### Life Context AI-generated monthly insights from your chat history. MemoryPlugin reviews your conversations and surfaces recurring themes, projects, and interests in a brief, readable summary. Available for Pro users. Shows up on your dashboard and is shared with MCP-connected AI tools for richer, more personalized context. ### Connected Apps New section in Settings to manage your connected applications: * See which apps have access to your MemoryPlugin account * Check when each connection was last used * Revoke access with one click ### Extension improvements * **Smooth animations** — polished transitions when opening and closing the panel * **Sync progress** — see exactly how many conversations have been synced during a batch * **Edit & delete memories** directly from the extension panel (v3.0.5) * **Memory load limit** — control how many memories are loaded at once (v3.0.6) * **Better mobile support** — improved touch interactions ### Bucket sorting Drag-and-drop reordering for your bucket list. Your preferred order is consistent across the dashboard and browser extension. ### Other improvements * Chat history upload limit raised from 500 MB to **1 GB** * Upload a single conversation file directly (ChatGPT or Claude export) * Improved memory injection — fewer cases where AI assistants refuse to use your context * Memories that failed to sync to search are now automatically repaired * Updated compatibility for Google AI Studio, Claude, ChatGPT, and Gemini * Multi-turn conversations in Ask — your full conversation history is preserved across messages * Browser extension versions 3.0.3 through 3.0.7 ### MemoryPlugin 2.0 A major update focused on simplicity and usability. The interface has been redesigned to reduce clutter, keep things closer, and help you stay in context. ### New sidebar layout Your buckets now live in a sidebar, always visible and one click away. No more hunting through menus. ### Unified Ask tool Talk to your memories, chat history, and files from one place. Rebuilt from scratch with: * **Conversation history** — pick up where you left off * **Referenced sources** — see which chats and files informed the answer * **Three modes** — Chat History, Memories, Files * **Copy responses** with one click Related docs: [Ask](/features/ask) ### File buckets Upload PDFs, Word docs, markdown, and text files. Query them through the Ask tool, MCP, or TypingMind plugin. Related docs: [File buckets](/features/file-buckets) ### Keyboard shortcuts * **⌘K / Ctrl+K** — Quick search chat history * **⌘J / Ctrl+J** — Add memory without leaving your page * **⌘Shift+O / Ctrl+Shift+O** — Jump to Ask tool ### Memory management * **Bulk add** — paste multiple memories at once, separated by line breaks * **Import/Export** — bring memories in from CSV or JSON, or export them out Related docs: [Import & Export](/features/import-export) ### UI improvements * **Clutter removal** — simplified, less busy interface throughout * **Dialogs and panels** — settings, add memory, and other actions open without losing context * **Inline suggestions** — review memory improvements without leaving your list * **Smart Categories easier to use** — redesigned panel with clearer controls * **Redesigned imports page** — clearer status at a glance * **Better copy** — clearer labels and explanations throughout ### Other additions * **Interactive setup guide** — floating checklist for new users * **Redesigned Get to Know You** — full-page chat experience * **Custom profile** — set a display name and avatar * **Roadmap & feedback** — accessible directly from the app *** ### Browser extension v3.0.0 * **Draggable floating button** — position it where you want, remembers per site * **TypingMind auto-sync toggle** — enable or disable in extension settings * **TypingMind sync performance** — batch loading handles thousands of chats much faster Related docs: [Browser extension](/integrations/browser-extension) *** ### TypingMind plugin * **Chat history support** — recall past conversations with `recall_chat_history` * **File search** — query uploaded documents with `search_uploaded_files` Related docs: [TypingMind plugin](/integrations/typingmind-plugin) *** ### MCP server improvements * **Better tool descriptions** — AI tools now understand when and how to use each tool more effectively * **New tools** — `get_full_conversation` (complete transcripts), `search_uploaded_files` (query documents) * **Renamed tools** for clarity: `search_memories`, `recall_chat_history`, `list_bucket_categories`, `list_category_memories` Related docs: [MCP server](/integrations/mcp-server), [Remote MCP server](/integrations/remote-mcp-server) ### Smart Memory enhancements * **Create custom categories** — define your own smart categories instead of relying only on auto-generated ones * **Remove categorization** — uncategorize buckets when you want to reorganize or start fresh ### Safari extension fixes * **Fixed chat history sync** — ChatGPT, Claude, and Gemini conversations now sync properly on Safari * **See which account is active** — user email now displayed in the extension panel * **Improved reliability** — better authentication and error handling for Safari users Related docs: [Browser extension](/integrations/browser-extension) ### Chat history improvements * **Protected pins** — pinned conversations can't be accidentally deleted * **Better search highlighting** — easier to see which parts of conversations matched your search * **Faster MCP queries** — Remote MCP server now runs multiple queries in parallel for quicker results Related docs: [Remote MCP server](/integrations/remote-mcp-server) ### Dashboard * **Quick help access** — contextual help links throughout the dashboard * **AskLibrary Plus bundle** — now included in Believer tier ### Extension compatibility * Updated for latest ChatGPT interface changes ### Background sync improvements * See sync frequency and the next scheduled run in Settings * More reliable provider switching with clearer controls * Safer limits when adding chat history context Related docs: [Browser extension](/integrations/browser-extension) ### Chat history relevance Smarter selection of what to include from past conversations, leading to more on‑topic answers. ### Chat history enhancements * **Pin important conversations** — keep your most valuable chats at the top for quick access * **Improved branch handling** — active conversation path is now correctly isolated, preventing messages from different branches from getting mixed together * **Modern table redesign** — cleaner, more refined interface for browsing your chat history * **Parallel queries** — Remote MCP inject tool runs queries in parallel so 10 queries take only as long as the slowest one, not 10x the time ### Expanding Core plan access Laying the groundwork to bring chat history features to more users: * Import limits and quota management for Core tier users * Clear usage visibility to help you track your chat history imports ### Performance improvements * Faster memory browsing with virtualized lists — smooth scrolling even with thousands of memories * In‑app search with Cmd/Ctrl+F — find memories instantly, with native browser search as fallback * Improved landing page with optimized hero video and better mobile experience ### Vision page New page sharing our long‑term vision and roadmap for AI memory. ### Reliability improvements and bug fixes * Fixed duplicate memory deletion handling * Better database error handling for chat history processing * Improved notification persistence — dismissed messages stay dismissed * Better mobile text sizing to prevent overflow on smaller screens ### Chat history (private beta) * Import conversations from ChatGPT, Claude, and TypingMind and search them in one place * Extension support to add chat‑history context to your messages * Available via Remote MCP as a chat‑history inject tool * New Chat History section in the extension settings * You can cancel adding chat‑history context while a message is sending Private beta — request access via Support. ### Chat with chat history (private beta) Ask questions about your imported conversations. Answers stream and include source excerpts for transparency. Private beta — request access via Support. ### New AI platforms * **Kimi (kimi.com)** support * **Z.ai (GLM)** support (reasoning blocks stripped for clean parsing) * **OpenRouter** support (streaming + message selectors) * **ChatLLM** platform support Learn more: [Browser extension platforms](/integrations/browser-extension) ### Enhanced search (Smart Memory + Query Mode, extension) * Less intrusive search overlay while browsing chats * Option to skip reranking for faster results * Better relevance for retrieved context ### Settings and sync (extension) * Redesigned Chat Sync tab with clearer status and controls * Settings menu overhaul for easier discovery (late September) Learn more: [Browser extension](/integrations/browser-extension) ### Remote MCP server (hosted) Connect from any MCP‑compatible tool over the internet — no local setup. Supports Smart Memory, basic memory tools, and a chat‑history inject tool. Learn more: [Remote MCP server](/integrations/remote-mcp-server) ### Memory suggestions Keep your memories dynamic and up to date: * Merge similar notes, remove duplicates, and flag outdated info * Evolve existing memories with fresher details instead of piling on new ones * Reduce clutter so relevant context surfaces more often * Review suggestions and apply changes with one click for full control ### Improved onboarding Helpful notifications and prompts to discover features and complete setup. ### Reliability improvements and bug fixes * Improved Smart Memory reliability with progress tracking and fewer stuck jobs * Fixed Claude issues in the extension ### Performance * Faster memory lists — pages open quicker and feel snappier * Quicker search — results update smoothly as you type * Better on mobile — navigation and actions feel lighter ### Pricing updates Introduced flexible monthly and annual billing options for the Pro plan with early-bird pricing for annual subscribers. ### Platform updates * **Perplexity**: New platform support via the browser extension * **DeepSeek**: Updated selectors for better compatibility * **Claude**: More resilient message interception on single-page navigation Related docs: [DeepSeek](/platforms/deepseek), [Claude](/platforms/claude) ### Memory Chat improvements Better dark‑mode styling and readability in Memory Chat. ### Dark mode (dashboard) Dark mode across core dashboard components for comfortable use at night. ### Settings page Better structure and mobile experience for Settings. ### Dashboard navigation New tabbed dashboard interface and a more compact action bar. ### Smart Memory Filter Smart Memory categories directly in the dashboard. ### Search relevance Better ranking for memory search results. ### Bulk actions Bulk delete memories directly from the dashboard. ### Feature Launch: Smart Memory Automatically organizes buckets (30+ memories) into meaningful categories with concise summaries. During a chat, MemoryPlugin can load only the categories that match the topic — reducing noise and token use while keeping answers focused. #### Benefits * More relevant answers by sending only what’s needed * Lower token usage and less clutter in long chats * Scales better as your memory collection grows #### How to use 1. Add 30+ memories to a bucket 2. Click “Create Smart Categories” in the dashboard — see the [Smart Memory guide](/features/smart-memory) 3. Refresh categories anytime as your memories evolve #### Post‑release reliability Smoother processing for large collections with clearer progress feedback and fewer interruptions. Related docs: [Smart Memory](/features/smart-memory) ### Saving reliability More dependable saving — memories save even if a selected bucket changes or isn't available. Automatically falls back to "General" bucket when needed. ### Browser extension improvements * Clearer text insertion with better formatting — memories appear properly separated and easier to read * More reliable memory loading across Claude, ChatGPT, Gemini, Grok, and AI Studio * Better timing when inserting context prevents text from merging together ### Sign‑in Added one‑time passcode (OTP) support for a smoother, more secure login. ### In development Foundational work on Smart Memory categorization system, preparing for June launch. ### Token usage tracking See how much context your buckets use with visual percentage indicators and color-coded warnings to help you stay within model limits. ### ChatGPT Canvas compatibility MemoryPlugin button now adjusts position automatically when Canvas opens, keeping controls accessible while you work. ### Support & feedback * Integrated live chat support for faster help * New feedback system to share suggestions and report issues ### Reliability improvements Various backend improvements and bug fixes for better stability. Learn more: [Browser extension](/integrations/browser-extension) ### New platform support * **Grok (x.ai)** - Access your memories while chatting with X's AI assistant * **Poe** - MemoryPlugin now works with Poe's multi-model platform Learn more: [Grok](/platforms/grok), [Browser extension](/integrations/browser-extension) ### Memory chat (beta) Ask questions about your stored memories and get AI-powered answers based on your memory collection. ### Text handling improvements Smoother memory insertion and formatting in chats, with better cursor behavior across platforms. ### Browser extension reliability Better bucket switching and improved ChatGPT message detection for more consistent behavior. ### Qwen support Added support for Qwen, expanding MemoryPlugin's reach to more AI platforms and giving you consistent memory access wherever you chat. Learn more: [Qwen](/platforms/qwen) ### Reliability improvements * Fixed Safari cookie handling for better browser compatibility * Improved bucket and profile initialization for new users ### Mistral support MemoryPlugin now supports Mistral AI! Access your memories while chatting with Mistral's powerful language models. ### DeepSeek support Get long-term memory with DeepSeek via the browser extension! Your memories work across all platforms - no lock-in, no starting from scratch with each new tool. Learn more: [DeepSeek](/platforms/deepseek) ### Team functionality (private beta) Initial team collaboration features for TypingMind custom deployments. Related docs: [Teams (TypingMind custom)](/teams/typingmind-custom) ### Interactive onboarding Smart onboarding that learns about you through conversation and automatically creates your first memories based on what you share — getting you set up with personalized context from day one. ### Bulk operations Perform operations on multiple memories at once: * Delete multiple memories in one action * Move memories across buckets in bulk * No more one operation at a time Learn more: [Bulk operations](/features/bulk-operations) ### Performance improvements Faster memory operations — adding, updating, and deleting memories now happen in the background so you don't have to wait. ### ChatGPT saved memories import Import your ChatGPT "saved memories" via the browser extension. ### Google AI Studio support Added compatibility for Google AI Studio platform. ### TypingMind browser extension support Access your memories directly in TypingMind through the browser extension — no plugin configuration needed. ### Self-hosted AI platform support Extension permission improvements enable compatibility with self-hosted AI platforms and custom deployments. ### Browser extension improvements Extension can now save multiple memories from a single AI response when multiple save commands are detected. ### API enhancements Bucket creation now available in the OpenAPI spec for custom integrations and automations. ### Performance improvements Faster memory moves — moving memories between buckets now happens quicker with parallel database updates. ### Reliability improvements Fixed duplicate memory detection to prevent accidental re-saves. ### iOS Safari support MemoryPlugin now works on iOS! Use the browser extension with Safari on your iPhone and iPad. Enjoy the same seamless experience across all your favorite AI tools from your iOS devices. The extension works exactly the same way as it does on Chrome on desktop. ### TestFlight access Early access available through [TestFlight](https://testflight.apple.com/join/3nBR4xdN) - install and start testing on iOS. ### Browser extension improvements * Faster memory loading with caching — your memories appear instantly when you reopen the panel * Visual feedback when saving memories — button shows spinner and checkmark states * Fixed issues with ChatGPT compatibility Learn more: [Browser extension](/integrations/browser-extension) ### Memory buckets Organize your memories like never before! Group related memories together with custom buckets: * Create "Personal" vs "Work" buckets * Separate "Project 1" from "Project 2" * Isolate related memories from unrelated ones * Have more control over which memories appear in each chat All memories default to a "General" bucket. Learn more: [Memory buckets](/features/memory-buckets) ### Browser extension updates Pick which bucket to save to when the AI suggests a memory — no need to move it later. ### Improvements * Won't save the same memory twice * Better memory insertion in ChatGPT ### ChatGPT support Full browser extension support for ChatGPT — access and inject your memories directly while you chat. Learn more: [ChatGPT](/platforms/chatgpt) ### MemoryPlugin Custom GPT Use MemoryPlugin inside ChatGPT with our Custom GPT. It can fetch and use your memories during a conversation — including an option to load your entire set when you ask. Learn more: [Custom GPT](/integrations/custom-gpt) ### Gemini support Google Gemini is now supported via the browser extension. Learn more: [Gemini](/platforms/gemini) ### LibreChat support MemoryPlugin now works with LibreChat so you can use memories on your self‑hosted setup. ### Claude improvements More reliable detection of new replies so save prompts appear at the right time, and memory insertion is more consistent. ### Dashboard * One‑click export: copy all memories to your clipboard to paste into any tool or keep a portable backup * See your total memory count at a glance ### Browser extension * Save commands are case‑insensitive — “to=MemoryPlugin” works just like “to=memoryplugin” * Gemini: no more cursor jumps when inserting context; text lands where it should * Load more context: up to 150 recent memories can be pulled in when you need extra detail ### Custom GPT * Added a “Load all memories” option so your Custom GPT can fetch your full set on demand ### Sign‑in and reliability * Fewer re‑logins thanks to automatic session refresh * Clearer OAuth consent screen * Newest memories appear first in the dashboard and extension ### Clearer setup when your token is missing or invalid in TypingMind * Where you see it: MemoryPlugin now shows a plain‑English message right in TypingMind when your token isn't set or is incorrect * What to do: the message tells you to copy your token from [https://www.memoryplugin.com/dashboard](https://www.memoryplugin.com/dashboard) and paste it into the TypingMind plugin settings * Consistent across actions: the same guidance appears when reading or saving memories so TypingMind can surface it directly ### Reliability improvements * Fixed error when viewing memories for the first time ### Add memories directly from the dashboard * Create new memories without needing to chat with an AI * Your memories appear immediately in your collection ### Performance improvements * Improved memory loading performance ### Support and policies * Added refund policy page for transparency about our refund process * Fixed checkout issues that prevented some users from signing up ### MemoryPlugin launches Give your AI conversations long‑term memory and stop repeating yourself. ### TypingMind integration MemoryPlugin debuts with a TypingMind plugin, bringing persistent memory to your AI conversations. Learn more: [TypingMind plugin](/integrations/typingmind-plugin) ### Core features * Store and retrieve memories across conversations * Vector-based semantic search for relevant context * Secure memory storage with authentication * Real-time memory injection into AI chats # Ask Source: https://help.memoryplugin.com/features/ask Query your chat history, memories, and files from one interface You know you worked out the deployment plan with an AI a few weeks back, but which chat was it in? And did that detail end up in a memory or a file? Ask is the one tool in the dashboard that answers natural-language questions about everything you have stored in MemoryPlugin. You type a question, it searches the source you point it at, and it writes back an answer with links to the originals so you can check its work. Ask is a single interface with three modes. It is not the same as the **Ask** entry inside the Chat History section, which only ever searches your imported chats. This page covers the unified Ask tool, which can switch between chat history, memories, and files in one place. Open it from the sidebar or press **⇧⌘O / Shift+Ctrl+O**. The Ask interface with History, Files, and Memories mode pills, a question about the launch date, and a streamed answer citing three source conversations The Ask interface with History, Files, and Memories mode pills, a question about the launch date, and a streamed answer citing three source conversations ## The three modes A pill toggle at the bottom of the composer switches Ask between three sources: | Mode | Searches | Use it to | | - | - | - | | **History** | Your imported and synced AI conversations | Find past decisions, project context, or how you discussed something before | | **Files** | Documents in your file buckets | Pull answers out of PDFs, Word docs, Markdown, and text files | | **Memories** | Memories in your memory buckets | Recall preferences, facts, and notes you have saved | The pills read **History**, **Files**, and **Memories**. Once a conversation has messages, its mode locks. To search a different source, start a new conversation. ## Using Ask Click **Ask** in the sidebar, or press **⇧⌘O** (Mac) / **Shift+Ctrl+O** (Windows/Linux). Choose **History**, **Files**, or **Memories** from the pill toggle. In Files and Memories mode a dropdown lets you scope the search. Leave it on **All files** or **All memories** to search everything, or pick one bucket for focused results. Type a plain-language question, or click one of the example chips under **Try asking:**. Press Enter to send. The answer streams in. Every answer cites the chats, memories, or files it drew from. Click a source to open the original.