Find it by what it looks like
Describe the shot and get matches, no tags required. Overexposed, shot from below, too much orange. This is the one your agent has that a filter panel doesn't.
NEW – Now in Claude, ChatGPT, and Cursor
Your agents can now do more
than just looking for your files.
Playbook speaks the Model Context Protocol (MCP), so your AI agents can handle tedious tasks without asking you where things are. Search, tag, version, organize, share.
Quick setup
Connect your workspace to
Claude, ChatGPT, or Cursor
Pick the tool you already work in, paste one URL, and authorize once.
-
Copy the Playbook connector URL
You'll paste this into Claude in the next step.
https://mcp.playbook.com -
Add it in Claude → Connectors
Open Settings → Connectors, add a custom connector, and paste the URL.
-
Sign in and start
Authorize Playbook once, then ask Claude to find, tag, or organize anything in your library.
-
Copy the Playbook connector URL
The same URL — it works with any client that speaks MCP.
https://mcp.playbook.com -
Add it in ChatGPT → Apps
In Settings → Apps, you may need Developer Mode enabled to add custom MCP.
-
Sign in and start
Authorize Playbook once, then pull assets straight from your boards into the conversation.
-
Add Playbook to your MCP config
Drop the connector URL into .cursor/mcp.json, or install it from Cursor's MCP directory.
https://mcp.playbook.com -
Sign in and build
Authorize once, then call Playbook from Cursor's agent — or skip MCP and hit the REST API directly.
How does Playbook MCP work?
- You
Post house just dropped the Scene 4 delivery into the workspace. Help me sort clips into Scenes selections, stills into Images, anything branded into Collateral.
- Playbook
Reading 16 new assets against the project boards...
- Playbook
Eight clips into Scene selections, three stills into Images, five into Collateral. The two re-renders of shots went in as versions. Nothing overwritten.
- You
Attachment SS26_product_sheet.csv
Attached the SS26 product sheet. Match column A to the filenames and fill in the fields.
- Playbook
Reading 96 rows, matching against 1,284 assets...
- Playbook
Matched 1,190 by SKU and wrote collection, material, and room onto each. 94 had no SKU in the filename, so those are on a board for you to assign.
- You
The vendor went through the 'Retouch queue' board and left some comments. Can you summarize the changes and set status 'Approved' on the asset they actually signed off?
- Playbook
Reading 14 comments left in Retouch queue…
- Playbook
8 marked as 'Approved' and 2 marked for 'In progress' based on what they wrote. Five I couldn't call either way, so those are untouched and grouped for you.
- You
Client wants to see where we're heading with the campaign. Pull the a few latest art direction, teaser, and event work into a board I can share with them.
- Playbook
Reading 342 assets across the three campaign boards...
- Playbook
38 on a new board, split the same three ways as the originals. 9 in Art directions, the teaser, and 28 in PR + Events. Ready to share when you are.
- You
Half of this board is iterations that got uploaded as separate files instead of versions. Can you replace them?
- Playbook
Reading 8 assets, matching by scene and pass...
- Playbook
5 folded onto their originals as new versions. Every earlier pass is still there and any of them can roll back. Two stayed put since I couldn't match to anything.
Tools
Endless possibilities
once it's connected
Stop describing where a file lives. Ask for the file.
Filter across everything at once
Tags, status, custom fields, file type, uploader, all combined in a single query. The kind of question that takes a person six passes through a grid.
Read the whole library, not a page of it
Walks the full board tree in one pass, nested boards included. This is what lets an agent answer questions about four thousand files instead of the forty on screen.
Write the metadata back
Titles, tags, custom fields, board placement. Point at a product sheet, match SKUs, and the fields land on every asset that matched.
Change state on a thousand assets at once
One call, across boards, approved or not. The afternoon of clicking after a review round, collapsed.
Bring a hundred files in per call
Give it URLs and it fetches, routes to the right board, and applies titles and tags on the way in. A catalog arrives structured instead of arriving as a pile.
Search and read
-
ai_searchFind assets by what they look like. -
search_assetsFilter by tag, status, field, or type. -
list_assetsRead a whole library or board tree. -
get_assetsFull detail for up to a hundred assets. -
get_asset_previewsRender assets at the size you choose.
Boards
-
list_boardsThe board tree, at any depth. -
list_board_childrenThe child boards under one board. -
get_boardA board's title, description, and count. -
create_boardA new board, nested where you want it. -
update_boardRename a board or move it elsewhere. -
copy_assetsCopy assets into another board. -
move_assetsMove assets to a different board.
Metadata
-
update_assetWrite titles, tags, and fields onto an asset. -
update_asset_statusSet the status label on an asset. -
change_asset_tagsAdd and remove tags, nothing else. -
set_asset_approvalApprove up to a thousand assets at once. -
list_custom_fieldsEvery field in the workspace and its options. -
create_custom_fieldA new field and its options.
Versions
-
create_asset_versionAdd a new version of the same asset. -
list_asset_versionsThe full version history, newest first. -
revert_assetRoll an asset back to an earlier version.
Sharing
-
share_boardA link anyone can open, password optional. -
publish_boardA public page, with download controls. -
asset_permalinksAdd, read, or revoke permanent public URLs.
Comments and notes
-
get_commentsRead the thread on an asset or board. -
create_commentPost a comment. -
create_noteA sticky note, for a brief or a caption.
Uploads
-
upload_from_urlsBring in a hundred files per call, tagged.
Workspace
-
list_organizationsWhich workspaces the account can reach. -
list_membersWho's in the workspace.
Destructive
-
delete_assetRemove an asset. -
delete_boardRemove a board and everything in it. -
delete_custom_fieldRemove a field and all its options.
Examples
Ready to give it
a try for yourself?
Give me a tour of our workspace, then suggested how to structure it better. What should be organized by boards vs tags?
What you have and where it's piled up, followed by a proposed board structure.
Reorganize our workspace so boards are one thing and tags carry the rest. Show me the plan before you move anything.
A plan first, then boards built and assets moved once you approve.
Everything in [board] needs filing. Sort in into the right boards, tag it to match and leave anything ambiguous.
Sorted and tagged, with a short list for you to check
Our tags have drifted. Find near-duplicates like [tag 1] and [tag 2] then merge them into one.
One tag instead of three spellings, so search stops missing things.
Find everything in our workspace that looks like [warm interior light, soft shadow] and copy it on a new board to share.
Matches found by what the images look like, not just file names.
Frequently asked questions
MCP (the Model Context Protocol) is an open standard for connecting AI assistants to outside tools and data. An app publishes an MCP server, and any MCP client can use it. Playbook publishes one, so an assistant can work with your workspace directly instead of you copying files into a chat.
Any client that speaks MCP. The steps above cover Claude, ChatGPT and Cursor because those are the ones teams ask for most, but the same connector URL works anywhere else.
It connects as you and works inside your account's permissions: it reaches the boards and assets you can already reach, and nothing that hasn't been shared with you. It can also make changes, the same ones you can make yourself. Tagging, filing, versioning, approving, creating boards, sharing, and bringing in files from a URL.
Deleting is a soft delete: anything removed goes to your Playbook trash and can be restored. The assistant is also instructed to confirm with you before deleting anything. On plans with version history, replacing a file keeps the original as a version you can roll back to.
Yes. One connection reaches every workspace your account belongs to, and the assistant works in whichever one you name. If you don't say, it will ask.
No. Playbook does not use your assets to train models. Once an assistant is connected, what it does with the files it retrieves is governed by that tool's own data policy.
All of them, starting with Free. MCP requests count toward your plan's monthly API calls: 1K on Free, 20K on Pro, 200K on Team, 5M on Business. See the pricing page for the full breakdown.
MCP is for the assistant: it discovers what Playbook can do and calls it on your behalf, in conversation. The REST API is for software you write. Both reach the same workspace, and the developer docs cover the API.