Your rooms, inside your AI assistant.
Connect Claude, ChatGPT, or another AI assistant to your Pactto account. From a normal chat, the assistant can read the transcripts, comments, PDFs, and docs a session produced, then write notes, tables, docs, and review statuses back into the room. Nothing to install: you paste one URL into the assistant's settings. If you work from a terminal with Claude Code or Codex, the Pactto CLI covers the same ground.
Make sure you have a Pactto account
The MCP server works on the same rooms you use in the browser. Your assistant signs in as you, so it sees the rooms you own or belong to and nothing else.
Add the server to your assistant
In Claude, open Settings, go to Connectors, choose Add custom connector, and paste the URL below. In ChatGPT, use Settings, then Connectors. Any other assistant that supports remote MCP servers takes the same URL. When the assistant asks, sign in with your Pactto account to approve access.
https://live.pactto.com/mcpAsk about your rooms
Start a new chat with the connector turned on and ask for something a room already holds. Listing your rooms is a good first request, because it confirms the connection and shows the names you will use in later prompts.
List all of my Pactto roomsPrompts that put the room to work
Paste any of these into a chat with the Pactto connector on, and change the names to match your own rooms and files. The room doesn't have to be open and nobody has to be in it: whatever the assistant writes is waiting for the team the next time they open the room in the browser.
What did we decide in room Brainstorming in the last 5 meetings?List the open comments in room Launch Video that mention mePull the text from every PDF on the canvas and list the deadlines you find.Take the meeting notes doc on the canvas and lay out each section as a sticky note on a new board called SummaryOrganize this room: group the images into one board, the videos into another, and the PDFs into a third, then align and space everything so it reads left to rightThe assistant reads what is scattered, decides the grouping, and the room comes out tidy without anyone dragging a shape.
Write a Pactto Doc with today's action items for Demian and SherifEveryone in the room sees the doc appear without reloading.
Create a Tasks table with Task, Owner, and Status columns and add a row for each open item from the transcriptAdd a comment at 0:42 on the hero cut saying the logo comes in too late, and set the video to Needs changesThe 36 tools your assistant gets
You never call these by name. The assistant picks them from what you ask, and this list shows what it can and can't do on your behalf.
Rooms
list_roomsLists the rooms you own or are an active member of, most recent activity first.get_roomA room's overview: name, description, your permissions, participants with online status, canvas and comment counts, review statuses, and room instructions.create_roomCreates a new room owned by you, as the web app does.update_roomRenames a room or changes its description.duplicate_roomCopies a room, with its canvas, boards, and content, into a new private room you own. Runs in the background.delete_roomDeletes (archives) a room, as the web app does. It disappears for everyone.
Reading content
list_assetsLists the items on the canvas: images, videos, PDFs, links, Pactto Docs, boards, notes, and more, with optional positions and sizes.get_assetThe details of one item: file name, type, size, dimensions, duration, pages, and review status. Can include the text of docs, tables, and notes.search_roomFinds a word or phrase across a room's comments, replies, and canvas content.get_commentsLists comments in a room, canvas pins and Pactto Doc comments, newest first, with authors.get_transcriptTranscripts of the meetings held in a room, as timestamped lines labeled with the speaker, with paging and time filters.describe_imageDescribes an image in the room using Pactto's image model. Spends AI credits.describe_pdf_visualDescribes the visual content of a PDF: slides, charts, scanned pages, and layout. Spends AI credits.
Adding files
import_fileBrings an image, video, or PDF into the room from a public URL.add_linkPuts a web link (http or https) on the canvas.get_operation_statusShows the status of a background job started by duplicate_room or import_file.
Pactto Docs
create_docCreates a native, collaborative Pactto Doc on the canvas.append_to_docAdds up to 50,000 characters to the end of a Pactto Doc.replace_doc_contentReplaces all the content of a Pactto Doc. Asks for confirmation first.
Canvas content
create_notesAdds 1 to 50 sticky notes in one call.create_textAdds 1 to 50 text blocks.update_textReplaces the text of a sticky note or text block.create_shapesAdds 1 to 50 geometric shapes.create_linesDraws 1 to 20 freehand lines.create_arrowsConnects existing items with 1 to 50 arrows.create_tableCreates a table on the canvas.edit_tableApplies 1 to 100 changes to a table at once, all or nothing: set cells, add rows, and more.
Boards and layout
create_boardCreates a board, a container on the canvas, optionally with items, a vertical, horizontal, or grid layout, and a color.add_to_boardMoves 1 to 50 existing items into a board.update_itemsChanges 1 to 50 items at once: position, size, color, parent board, and more.align_itemsAligns 2 to 50 items to the top, middle, bottom, left, center, or right.distribute_itemsSpaces 3 to 50 items evenly, horizontally or vertically.delete_itemsDeletes 1 to 50 items. Asks for confirmation first, showing exactly what will be removed.
Review and feedback
add_commentAdds a comment as you: on the canvas, an item, a video frame or range, or a doc.add_annotationDraws freehand strokes tied to a comment pin.set_review_statusSets the review status of 1 to 50 items: Draft, In progress, Needs review, Needs changes, or Approved.
Permissions, credits, and what comes next
It acts as you. Every tool runs as the signed-in user, with the same role you have in each room. A viewer can read but not edit, and the assistant can't reach a room you aren't in.
Deletes ask first. delete_items, delete_room, and replace_doc_content change nothing on the first call. They return what would be removed, and the assistant has to come back with your confirmation.
Retries are safe. Write tools take an idempotency key, so a retried request never applies the same change twice.
Room content is data. Comments, transcripts, and docs are written by people. The assistant treats them as material to read, never as instructions to follow.
Credits. describe_image and describe_pdf_visual spend Pactto AI credits. Reading a PDF's text with get_asset is free, and so is every other tool.
Coming soon
- Uploading files from your computer, not only from a public URL
- Getting a room's guest link
- Inviting people to a room by email, and removing them
- Reading room chat messages
- Downloading files from a room
Put a room in your assistant.
Create an account, then paste the server URL into Claude or ChatGPT.