Connect Claude to Dropbox

Return the connected Dropbox account: its account ID, display name, email address, and whether the account belongs to a team. It touches no file, which makes it the cheapest proof that the connection works and the right call to confirm whose Dropbox this is before acting on it.

Return how many bytes this account has used against its allocation. Numbers only - no file, no folder and no name appears in the answer. Call it before a large copy to know whether there is room.

List the contents of one folder. Paths are Dropbox-style and start with a slash - "/Clients/Acme" - and the account's root is the empty string, which is what this sends when no path is given. The answer carries `cursor` and `has_more`; when `has_more` is true, pass that cursor to list_folder_continue for the next page rather than calling this again. Set recursive to walk every subfolder in one call, but be aware Dropbox warns that a deep tree can time out - walking one level at a time is the reliable way through a large account.

Fetch the next page of a listing started by list_folder, using the cursor that listing returned. The cursor also works as a change token: calling this again later returns only what has been added, changed or removed since, which is how a folder is watched without listing it repeatedly.

Search the account for files and folders by name, and by the text inside documents on paid plans. Returns matches with each item's path, so a hit can be handed straight to get_metadata or create_shared_link. Narrow it with search_path to look inside one folder, with file_extensions to keep to one kind of file, or with filename_only when the words are expected in the name rather than the contents.

Read one file's or folder's metadata by path or by ID - its name, size, revision, when it last changed, and whether it is shared. Dropbox does not support metadata for the account's root, so this needs a real path; use list_folder to see what is at the top level.

List the shared links this account has already created, with the file or folder each one points at. Give a path to see the links for one item; omit it to page through every link on the account. Call this before create_shared_link - Dropbox refuses to create a second link with the same settings for an item that already has one, and the existing link is usually the answer anyway.

Create a shared link for a file or folder. Classified destructive, alongside deleting: with audience "public" the link works for anybody who has the URL, signed in or not, and a URL that has been sent somewhere cannot be un-sent. Set audience "team" to keep it inside the Dropbox team, or "no_one" for a link that only points members at content they can already reach. Prefer list_shared_links first - Dropbox rejects a second link with the same settings on an item that already has one.

Delete a file or folder. Deleting a folder takes everything inside it. Dropbox keeps a deleted item recoverable from the web interface for the account's retention window - 30 days on most plans - but it leaves the folder now, which is why this is gated with sharing rather than with renaming. This is not permanent deletion: that needs the files.permanent_delete scope, which this connector never asks for, so no grant here can destroy something outright.

Move or rename a file or folder. Classified destructive rather than write because the old location is gone afterwards: anybody or anything holding the previous path - a link in a document, a script, a colleague's bookmark - no longer finds it, and there is no undo. Use copy_file when the original should stay where it is.