Skip to content
Qind
Start free
All connections

Use Qind directly from Claude and ChatGPT.

Qind speaks the open Model Context Protocol, so your AI assistant can answer questions from your library with citations, read long documents and transcripts, manage tasks and reminders, and save notes, bookmarks and uploads without leaving the tool you already use.

Setup steps below for Claude Code, the Claude app, ChatGPT, Cursor, Codex and Gemini CLI.

Add Qind to your AI See setup steps

Everything Qind does, now as tools your AI can call.

35 tools: ask and read, organize, capture and upload, tasks and reminders, digests, and AI actions on an item. Every call acts as you, on your own library.

  • Ask and read

    • Ask a question and get an answer that cites your saved items ask_qind
    • Search your library by meaning and keywords search_items
    • See what is in a collection list_collection_items
    • Read an item's details get_item
    • Read long documents page by page, plus OCR text and timestamped transcripts get_item_content
    • See the stored summary, takeaways, topics, captions and key moments get_item_insights
    • Read a note as Markdown get_note_markdown

  • Organize

    • List your collections get_collections
    • List your tags list_tags
    • Add, remove or replace an item's tags set_item_tags
    • Move an item to another folder move_item
    • Rename an item or a folder rename_item
    • Delete an item delete_item

  • Capture and upload

    • Write a new note create_note
    • Edit an existing note edit_note
    • Save a link as a bookmark create_bookmark
    • Save selected text with its source save_clip
    • Start uploading a file, up to 100 MB, from a client that can make an HTTP upload prepare_upload
    • Finish an upload complete_upload
    • Check whether an upload is ready get_upload_status

  • Tasks and reminders

    • List your tasks, linked to the item they came from list_tasks
    • Create a task create_task
    • Change a task update_task
    • Complete or reopen a task complete_task
    • List your upcoming reminders list_reminders
    • Set a reminder in your time zone create_reminder
    • Reschedule a reminder update_reminder
    • Cancel a reminder cancel_reminder

  • Digests and suggestions

    • List your weekly and monthly digests list_digests
    • Read a digest get_digest
    • Get items to revisit, related items and open tasks, from your own data get_ai_recommendations

  • AI actions on an item

    • Summarize an item summarize_item
    • See where Qind proposes to file an item; you confirm in the Qind app classify_item
    • Find the commitments in a meeting, note or document; creates no tasks extract_action_items
    • Create only the tasks you select, linked to the source item create_tasks_from_item

Save to Qind from any conversation.

Not a mockup. Here's the actual round-trip: ask your AI to save the conversation, and Qind files it as a searchable note in your library.

ChatGPT calling the Qind Create note tool after the user asks to add the chat to Qind
In ChatGPT. "add this chat to Qind" → create_note tool runs.
The saved note open in Qind with Ready status
In Qind. The note lands in your library, ready to search.

You decide what your AI can touch.

Your client signs in to Qind through your browser, and every call acts as you. Nothing connects until you approve it.

Qind OAuth authorization screen for Claude Code with approve and deny access buttons

15 tools only read

Searching, reading items and transcripts, listing tasks and reading digests change nothing in your library.

20 tools make changes

Notes, bookmarks, uploads, tags, tasks and reminders. Two of them delete or cancel: delete_item and cancel_reminder. Each tool tells your client which kind it is, so a client can ask you first.

  • Your library only. No tool takes a user id: each call acts as the person who signed in, so a client cannot reach anyone else’s library.

  • Suspicious text is flagged. Saved pages can contain instructions aimed at an AI. When Qind returns text that looks like one, it marks it for your client.

  • Plan limits are named. A call that would pass a plan limit fails with a code that says which, such as ITEM_LIMIT_REACHED.

  • No training by Qind. Qind does not train models on your content. What your AI reads is sent to that AI’s provider, under its terms.

Add Qind to your MCP client.

New to Qind? MCP connects to your existing library, so create a free account first, then connect any client below.

Installation

or follow the steps, checked against Claude Code’s guide
  1. Add the server from your terminal.
    claude mcp add --transport http qind https://mcp.qind.ai/mcp
  2. Start Claude Code and run /mcp.
  3. Choose qind and sign in to Qind in the browser window that opens.

Every client connects to Qind's hosted server at https://mcp.qind.ai/mcp and signs in with OAuth in your browser. There is nothing to install or run locally.

Other clients. VS Code, Windsurf, Cline and Zed can connect to remote MCP servers, but their sign-in with Qind is not tested yet. A client that supports remote servers over Streamable HTTP, with OAuth sign-in and dynamic client registration, can use the same URL.

Five things to ask first.

Each one uses tools from the list above. Your client decides which to call; these are the ones it usually picks.

  1. “Search my Qind library for what I saved about pricing and tell me what we decided.”

    search_itemsget_item_insightsask_qind

  2. “Save this conversation to Qind as a note called 'Launch plan decisions' in my Product research collection.”

    get_collectionscreate_note

  3. “Read the transcript of my last customer call and list the action items. I will pick which become tasks.”

    search_itemsget_item_contentextract_action_itemscreate_tasks_from_item

  4. “What is due this week? Remind me tomorrow at 9 about the contract task.”

    list_taskscreate_reminder

  5. “Tag everything I saved about pricing with 'pricing', and tell me where Qind suggests filing my newest bookmark.”

    search_itemsset_item_tagsclassify_item

Questions about MCP and Qind.

Which clients can upload files to Qind?
Claude Code, Codex and Cursor can: an upload is an HTTP request the client makes outside the chat, up to 100 MB. The Claude app and ChatGPT cannot, so save a link or paste the text instead.
Does asking through MCP use my plan?
ask_qind uses one chat message from the same allowance as asking in the Qind app. Plan limits are on the pricing page.
What does my AI provider see?
What a tool returns: the search results, notes and text your AI reads go to that AI’s provider, under its terms. Qind does not train models on your content.
How long does ask_qind take?
It waits up to a minute for the answer. If the answer is not ready by then, it says the run is still going; asking again starts a new run.
Can my AI delete things?
Two tools delete or cancel: delete_item and cancel_reminder. Each tool tells your client whether it only reads, writes or deletes, so a client can ask you before it acts.

Building an agent? The Qind for AI agents page lists every tool, the sign-in flow, errors and limits in Markdown.

Capture from wherever knowledge appears.

Your saved knowledge, available where you ask.

Connect Qind once. Then search, capture, and organize from the AI client you already keep open.

Create a free account Add Qind to your AI

Free plan, no card needed. Upgrade only if you need more.