Beta
Sign in

Features

What MindDuck can do today

Everything that already works, by group. Every item has an article in the documentation

Chats and memory

  • Any model answers with memory across all your conversations
  • Facts with a source and a date: you confirm them
  • Conversation topics and a topic map
  • Long chats compress into a summary, nothing gets lost
  • Source cards under the answer, with a quote and a date
  • Personal instructions you can change right from the chat
More in the documentation

Search

  • By meaning, exact and hybrid, offline on the device
  • Filters by chats, documents, facts and dates
  • Search by day and inside a single chat
  • Images are found by their description and the text on them
More in the documentation

Import

  • ChatGPT and Claude archives with their original dates
  • A fresh archive loads without duplicates
  • Claude Code and Codex sessions become chats
  • TXT, Markdown, Word, Excel and PDF files as documents
More in the documentation

Models

  • OpenAI, Anthropic, Gemini, DeepSeek, Mistral, OpenRouter with your own key
  • Local Ollama and LM Studio without the internet
  • Switch models for any chat and for every request
  • Pay the provider directly, no subscription
More in the documentation

The web

  • The model's built-in search or one of your choice - Wikipedia for free, Brave, Tavily, Exa
  • Pages are read on the device in a background browser
  • Long pages in parts, with a table of contents
  • A snapshot of the page for the model, only with your permission
  • Images saved on request into the chat and onto the device
More in the documentation

Files and folders on the device

  • Project folders: the assistant reads text, Word, Excel and PDF in place
  • Files are not copied and never leave the device
  • Changes only with your confirmation, with undo
  • On the phone and tablet through the system folder picker
More in the documentation

Projects and documents

  • A project gathers chats, documents, facts and folders with a description
  • The assistant reads and edits documents, projects and facts on request
  • An editor with versions, rollback and a draft on the device
  • Export to Markdown, TXT, PDF and HTML
  • A 30-day trash
More in the documentation

Voice and attachments

  • Dictation through OpenRouter, OpenAI, Groq or Mistral
  • Auto-send after recognition
  • Up to 8 images and 4 PDFs in one message
  • Image descriptions and text for search
More in the documentation

Memory for your programs

  • Claude Code, Codex, Claude Desktop and others work with MindDuck memory
  • Only in the projects you choose and only on this computer
  • Project context at the start of every session, working notes
  • A Chrome and Firefox extension: save a page or discuss it
  • Terminal commands and a local HTTP API
More in the documentation

MCP connections

  • Any MCP server: paste JSON, point to a program or an address
  • Every action with permissions and confirmation
  • Keys and OAuth are stored encrypted

Privacy

  • AES-256-GCM encryption, the key is only yours
  • Private chats and documents that stay out of memory
  • The sync cloud sees only encrypted packets
  • MindDuck has no server with your data
More in the documentation

Devices and cloud

  • Windows, macOS, Linux, Android
  • One base on all your devices, a new one joins with a QR code
  • An encrypted copy in a cloud of your choice, restored with your code
  • A backup as a file
More in the documentation

In the works

What comes next

Only what we already expect to build

  • iOS: the same app and the same memory on iPhone and iPad
  • Sync through Dropbox, OneDrive and WebDAV
  • Document and chat side by side: discussing and editing text together with AI, including a selected passage
  • A voice assistant: a live conversation, a quiet mode, a hard question for a strong model right in the conversation
  • App lock with a PIN or biometrics
  • Quick capture: a global hotkey and a tray icon
  • Automatic updates
  • The interface in more languages
  • The app as a memory system for agents - Claude Code, Codex and so on

Maybe later: image generation and optional MindDuck services for those who do not want to set anything up. Keeping your data with you and using your own key will remain a full way of working

Let's build the right personal AI assistant together!

Start keeping what matters today

Join the beta for free