DocSend

Use the docsend CLI to download standalone documents and files from DocSend Spaces. Progress goes to stderr and every command prints JSON to stdout.

Before starting, check whether the thread already contains the requested file, an attachment_ref, or a usable upload. If it does, use that file instead of opening DocSend again.

Choose The Command From The URL

• If the URL path contains /view/s/, it is a DocSend Space. Start with docsend login.
• For any other DocSend view URL, it is a standalone document. Start with docsend download.

Do not run download first for a Space. Do not run login for a standalone document.

Standalone Documents

Use download for a regular DocSend URL:

docsend download '<docsend-url>' \
  --email '<email>' \
  --output '<output.pdf>'

Add --passcode '<passcode>' only when the user supplied the document passcode. Use --session-timeout 1800 on the initial command when the verification handoff may need the maximum supported time.

On status: ok, use the file at the returned output path.

Email Verification

When download or login returns status: verification_link_required:

1. Preserve resume_session_id and expires_at from the JSON response.
2. Obtain the newest matching DocSend verification URL using the Gmail workflow below or another authorized email tool. Extract it without opening it.
3. Resume the same session before it expires:

DOCSEND_VERIFICATION_LINK='<full-verification-url>' \
  docsend resume '<resume-session-id>' \
  --output '<output.pdf>'

Omit --output when resuming a Space. A successful Space resume returns status: deal_room_ready and its root inventory.

If the link is invalid, retry resume with the correct link and the same session ID. If the session expired, start one new download or login command. Never open a verification URL outside docsend resume and never start duplicate sessions while one is active.

Find The Link In Gmail

When gsuite is available, search the Gmail account connected to the current Centaur user. Gmail operations use userId=me; never search another mailbox or ask for Gmail credentials.

1. Search for recent DocSend mail and inspect results newest first:

centaur-tools call gsuite gmail_search \
  '{"query":"newer_than:1d from:docsend.com","max_results":5}'

2. Read the newest candidate received after the current verification request:

centaur-tools call gsuite gmail_get '{"message_id":"<message-id>"}'

3. Extract the full HTTPS DocSend verification URL from the message body or HTML. Confirm the hostname is docsend.com or a subdomain. Do not click it, request it, preview it, or pass it to any browser or HTTP tool.
4. If no matching message has arrived, repeat the search up to three times at roughly 10-second intervals, always checking the newest results first. Do not submit the DocSend email gate again while polling.
5. Pass the extracted URL through DOCSEND_VERIFICATION_LINK to docsend resume, as shown above. Do not put verification URLs in command arguments or logs. The resume command is the only operation allowed to open it.

If Gmail is unavailable, unauthorized, or still has no message after the bounded polling, preserve the session details and use another authorized email tool or ask the user for the link. Do not start another Browserbase session while the existing one remains valid.

Spaces

Start a /view/s/ URL with login:

docsend login '<docsend-space-url>' --email '<email>'

For a larger bulk download, add --session-timeout 1800 to the initial login command so the agent has time to traverse folders and fetch each document.

After login or verification returns deal_room_ready, keep its session_id for every following command.

List the Space root:

docsend list '<session-id>'

Open a folder using its item ID:

docsend open-folder '<session-id>' --folder-id '<folder-item-id>'

Both commands return only one directory level. Read each folder's ID from the preceding JSON response and call open-folder again to traverse deeper.

For bulk downloads, inspect the returned items, traverse any folders, and call fetch once for each item whose type is file. Fetch files sequentially so the agent can choose output paths and report an individual failure without losing the rest of the inventory. Treat every fetch as independent: if one item does not return status: ok, record its path, status, and error, then move on to the next file. Do not stop the bulk download because one item failed. This is common for unsupported file types such as videos.

docsend fetch '<session-id>' \
  --item-id '<file-item-id>' \
  --output '<output-path>'

When the owner enabled downloads, fetch uses DocSend's regular download button and returns the original file with download_method: original. When the button is disabled, fetch opens the document viewer in the authenticated Space session and uses the standalone rendered-page recovery to return a PDF with download_method: rendered_pdf. The downloaded filename may differ from the visible Space title. Do not pass folder or external URL item IDs to fetch.

Close the session when finished:

docsend close '<session-id>'

Handling Results

• Parse stdout as JSON even when the command exits nonzero.
• For email_required, ask for an email unless the user already supplied one.
• For passcode_required, ask only for the document passcode.
• Preserve session fields on verification and failure responses.
• During a bulk download, continue to the next file after any per-item failure and summarize all skipped or failed items after attempting the full inventory.
• Report statuses such as blocked, expired, item_not_found, not_downloadable, and download_unavailable exactly. Only recover pages that the authenticated DocSend viewer renders; do not bypass access controls.
• Do not expose email addresses, passcodes, verification URLs, API keys, or downloaded file data in the final response.