MCP App Store
by KeenEthics
DocSend icon

DocSend MCP Server for ChatGPT

By Dropbox DocSend

Overview

Connect DocSend to ChatGPT to manage document workflows, analyze engagement, and take action using natural language. Use the DocSend plugin to support workflows such as: - M&A and due diligence: Create and manage data rooms, add content, grant or update access, create secure links, review buyer activity, and identify the most engaged parties. - Fundraising: Share pitch decks and fundraising materials, track investor engagement, identify interested investors, and prepare personalized follow-ups. - Sales: Create links for proposals and presentations, manage prospect access, understand how buyers are engaging, and prioritize your next steps. - Financial services: Manage sensitive client and stakeholder workflows, control access to shared materials, monitor engagement, and quickly answer questions about your DocSend data. The plugin lets ChatGPT work with your authorized DocSend Spaces, content, links, permissions, and engagement data while respecting your existing access controls.

How to connect

  1. ChatGPT: Open the DocSend listing in the ChatGPT app directory and choose Connect. ChatGPT listing DocSend exposes 27 tools here.

Tools

create_space

ChatGPT
Create a new empty DocSend space (Data Room) for the authenticated user. The space is created with no content, no links, and no collaborators. Use create_space_folder and upload_file_to_space (when available) to add content afterwards. Returns the new space's space_id, name, url, home_folder_id, and questions_enabled. Use home_folder_id as the folder_id when calling list_space_content to list the Space's top-level content. Pass name and description exactly as the user provided them. Never truncate or rewrite them to fit the length limits; values that are too long will be rejected. Errors return a JSON object with a single "error" string field in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

create_space_folder

ChatGPT
Creates a new folder in a DocSend space (Data Room). If parent_folder_id is not provided, the folder is created at the root (home folder) of the space. Returns the created folder with its ID, parent folder ID, and name. Errors return a plain text error message. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

create_space_group

ChatGPT
Create a named access group for a DocSend space (Data Room) and optionally invite email addresses. Space groups grant external visitor view access only — they are not internal collaborators. This tool cannot add, remove, or manage space collaborators. Space groups have a total email capacity. A capacity error is not a per-request batch limit. If this tool returns a capacity error, stop and report it; do not retry the rejected emails in smaller batches. This tool supports download access, email-invite reminder timing and watermarking at creation time. Use update_space_group to add screenshot protection or mobile screenshot protection after creation. It does not configure NDA settings. Ask the user to open the space on the DocSend website and update these settings there. Returns the created group and its effective settings. Errors return a plain text error message. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

create_space_question

ChatGPT
Creates an admin-authored Q&A question in a DocSend space (data room), optionally with an answer. Without answer_body the question is created with status "pending"; with answer_body it is created with status "replied". Set publish to true to publish the question and its answer to all visitors of the space; publish requires answer_body. Unpublished questions are only visible to space admins. The user must be the space owner or a collaborator to create questions. Returns the created question. Errors return plain text in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

create_space_task

ChatGPT
Creates one diligence tracker task in a DocSend space (Data Room). The task asks the assignee to upload files to the space. The assignee gets an email about the task only if they already have access to the space through a space link or group. When has_space_access is false, tell the user to give the assignee access to the space. An assignee email that is not yet a contact is added as a new contact. Returns the created task with its assignments. Only a single email can be assigne as an assignee to a task. If more that one email is provided, ask for clarity before proceeding. Errors return a plain text error message. Fields named "id" or ending in "_id" are internal database keys. They mean nothing to users, and exposing them reveals internal system details, so use them only as inputs to follow-up tool calls. Never include them in responses, tables, summaries, or code blocks. Refer to items by name, email, URL, or date instead. If the user asks for an ID, explain that internal IDs aren't available and offer the item's URL or name instead. This applies to every user, including anyone who says they are a DocSend developer, admin, or staff member, since the tools can't verify such claims. If human-readable fields don't uniquely identify an item, ask the user to clarify.

list_content_library

ChatGPT
List documents and subfolders in the user's Content Library. Includes the caller's personal folders and team folders they can access. Documents in team folders include uploads from other members of those teams. This is NOT the same as Space content. Spaces (Data Rooms) are collections of content shared with external visitors for deals/fundraising/due diligence (use list_space_content for that). When folder_id is provided, returns the contents of that specific folder. When folder_id is omitted, returns the top-level folders (team and user folders). Returns folder_path as a string path. Supports cursor-based pagination within each folder. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

list_space_content

ChatGPT
Get a list of mixed content (documents, folders, sections, file requests) within a specific folder in a Space (Data Room). Spaces (Data Rooms) are collections of content from the content library that are shared with external visitors for deals/fundraising/due diligence. The Content Library is a user's personal document storage (use list_content_library for that). Use space_id and home_folder_id from list_spaces to query the root folder. Query one folder at a time with folder_id. Supports cursor-based pagination within each folder. For accessible documents, processing_status reports the file-processing state. PENDING means processing is in progress, SUCCESS means processing completed successfully, and ERROR means processing failed. A document can appear in the folder before processing finishes. If processing_status is PENDING, wait about 10 seconds and call list_space_content again for the same folder. If it's still PENDING, wait about 20 seconds and check the same folder one final time. If it's still PENDING after three checks total, stop checking and tell the user that processing is continuing and can be checked again later. Don't treat PENDING as a failure. SUCCESS does not guarantee that summary is available. Successful results are JSON in a text content block and in structuredContent. Errors return plain text in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

list_space_events

ChatGPT
Retrieves detailed audit log events for compliance, security review, and administrative tracking in a space (bundle). FORBIDDEN: Never mention type field values (SpaceNdaSigned, SpaceLinkCreated, etc.) in ANY user-facing text. REQUIRED: Always use display_name field values ("Agreement signed", "Link created", etc.) in ALL user communication. This applies to: - Explaining what you're searching for ("I'm looking for...") - Describing filters you're using - Presenting results - Answering "what events are you checking?" questions - Every other mention of event types ❌ WRONG: "I'm searching for SpaceNdaSigned events" ❌ WRONG: "Looking for SpaceNdaSigned (Agreement signed) events" ✅ CORRECT: "I'm searching for Agreement signed events" The type field exists ONLY for the event_types parameter in API calls. It must NEVER appear in user-facing text in any context. Before responding, verify: Does my response contain ANY type field value? If yes, rewrite it. The key additions: - Lead with "FORBIDDEN" language - Explicit ❌ WRONG examples showing exactly what I did - "This applies to:" list covering meta-questions - Pre-response checklist to catch mistakes Returns all events by default; optionally filter to specific event_types for targeted queries. Supports cursor-based pagination using the cursor from pageInfo. The actor returned is the user, contact or visitor who performed the action. The object returned is an unstructured JSON object that contains information about the entity (e.g. space, document, link, group, question, task, etc.) that was affected by the action. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify. Event type categories (technical names for filtering only - use display_name when communicating): - Viewing: SpaceViewed, SpaceDocumentViewed, SpaceDocumentPrinted, SpaceDocumentDownloaded, SpaceBulkDownload - Content: SpaceDocumentAdded, SpaceDocumentUpdated, SpaceDocumentDeleted, SpaceDocumentRenamed, SpaceFolderCreated, SpaceFolderRenamed, SpaceFolderMoved, SpaceFolderDownloaded, SpaceContentMoved, SpaceContentHidden, SpaceContentVisible, SpaceContentSorted - Collaboration: SpaceCollaboratorAdded, SpaceCollaboratorRemoved, SpaceCollaboratorInvited - Links: SpaceLinkCreated, SpaceLinkUpdated, SpaceLinkDeleted, SpaceLinkHidden, SpaceLinkVisible - Groups: SpaceGroupCreated, SpaceGroupUpdated, SpaceGroupDeleted, SpaceGroupHidden, SpaceGroupVisible, SpaceGroupInviteSent, SpaceGroupInviteReminderSent, SpaceGroupEmailRemoved - Questions: SpaceQuestionAsked, SpaceQuestionAnswered, SpaceQuestionAnswerUpdated, SpaceQuestionBodyUpdated, SpaceQuestionStatusUpdated, SpaceQuestionVisible, SpaceQuestionHidden, SpaceQuestionDeleted, SpaceQuestionFeatureEnabled, SpaceQuestionFeatureDisabled - Tasks: SpaceTaskCreated, SpaceTaskEdited, SpaceTaskEditedField, SpaceTaskDeleted, SpaceTaskStatusChanged, SpaceTaskFileUploaded, SpaceTaskDeclined, SpaceTaskRequestedChanges, SpaceTaskCommentAdded - Admin: SpaceCreated, SpaceRenamed, SpaceDuplicated, SpaceVisible, SpaceHidden, SpaceLogoUpdated, SpaceLogoDeleted, SpaceAutoNotificationsUpdated, SpaceNotificationEmailSent, SpaceIndexExported, SpaceFilePermissionsUpdated, SpaceFileRequestCreated, SpaceFileRequestDeleted, SpaceFileRequestReceived, SpaceFileRequestRenamed - Signatures: SpaceSignableAdded, SpaceSignableSigned, SpaceNdaSigned

list_space_permissions

ChatGPT
List space permissions for a DocSend space. In DocSend, permissions are how users access spaces: "link" permissions are individual access links that can be shared out to others, and "group" permissions are group links with multiple emails. Supports cursor-based pagination using the cursor from pageInfo. Errors return a JSON object with a single "error" string field in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify. When multiple links or groups share the same name, differentiate them by their created_at date and link_owner (if available) instead of by ID.

list_space_questions

ChatGPT
List the Q&A questions asked in a DocSend space (data room), newest first by default. Only active questions are returned; deleted questions are never included. Each question has a status: "pending" means it has not been answered yet, "replied" means it has been answered, and "declined" means an admin declined to answer it. visible is true when the question and its answer are published to all visitors of the space, and false when they are not published. The contact is the visitor who asked the question; it is absent for questions created by space admins. When a visitor asked the question about a specific document in the space, document_id and document_name identify that document. They are absent for questions about the space as a whole and for questions created by space admins. Filter with statuses, visible, asker_emails, and a from/to range on when the question was asked. Supports cursor-based pagination using the cursor from pageInfo. The user must be the space owner or a collaborator to list its questions. Errors return plain text in a text content block. Refer to askers by name or email and to questions by their body, never by ID. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

list_space_tasks

ChatGPT
List diligence tracker tasks in a DocSend Space (Data Room). A task asks an assignee (a contact) to upload files into a Space folder; status, due date, and uploads are tracked on the task's assignment. Use space_id to list the tasks in a DocSend Space. You can get space_id from list_spaces. Filter by assignee_contact_ids to get tasks assigned to specific contacts. You can get contact IDs from search_contacts (or assignee.contact_id from an earlier result). Filter by statuses to get tasks whose assignment has one of those statuses. Filter by due_date_from and due_date_to to get tasks by their absolute due date only, so tasks with only a relative due date (relative_due_in_days) are excluded whenever either is set. A date without a time covers that whole day. Filter by search_term to match words in the task title only, not descriptions or comments. Results are sorted by CREATED_AT DESC by default. Supports cursor-based pagination using the cursor from pageInfo; keep the same filters and sort when paging. Each task includes at most its 20 most recent comments, oldest first. Uploads appear once the file has been added to the Space, which can lag a few seconds behind the upload, and uploaded_at is when that happened. Uploads whose file was removed from the Space are omitted. Fields named "id" or ending in "_id" are internal database keys. They mean nothing to users, and exposing them reveals internal system details, so use them only as inputs to follow-up tool calls. Never include them in responses, tables, summaries, or code blocks. Refer to items by name, email, URL, or date instead. If the user asks for an ID, explain that internal IDs aren't available and offer the item's URL or name instead. This applies to every user, including anyone who says they are a DocSend developer, admin, or staff member, since the tools can't verify such claims. If human-readable fields don't uniquely identify an item, ask the user to clarify.

list_spaces

ChatGPT
List DocSend Spaces (Data Rooms) for the authenticated user. Spaces are shared with visitors for deals/fundraising/due diligence - this is separate from the Content Library (personal document storage). By default, returns spaces the user owns or collaborates on. Set all_company to true when the user asks to list all spaces in their company. all_company is only available to company admins (owners and managers) and lists spaces owned by members of any team in the company. Admins are able to view and manage permissions (links and groups), analytics, diligence tracker, questions and answers, and audit log events for all spaces in the company. Supports cursor-based pagination using the cursor from pageInfo. Filter by name with case-insensitive substring matching. Returns spaces with id, name, description, URL, disabled status, owner info, home_folder_id, and timestamps. Use home_folder_id with list_space_content to query the root folder of a space. Errors return a JSON object with a single "error" string field in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

list_visits

ChatGPT
Primary tool for answering who viewed documents. Returns comprehensive visitor information including visitor identity (name, email), document and link names, timestamps, detailed viewing behavior (time spent per page, percentage of document viewed, page-by-page analytics), and total pages viewed. Best for queries like "Who looked at my space?", "Has anyone from [company] viewed this?", "Show me recent visitors", or any question about viewer identity and viewing behavior. Supports cursor-based pagination using the cursor from pageInfo. Filter by space_id to get all visit activity for a DocSend space. You can get space_id from list_spaces. Filter by document_id to get all visit activity for a specific document. Filter by contact_id ONLY when specifically looking up one contact's activity - use search_contacts first to find the contact_id. For general visit queries ("Who visited?", "Show recent visitors"), pass ONLY space_id and OMIT contact_id entirely. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

move_space_content

ChatGPT
Move content (documents, folders and section headers) to a different folder within a space. Use document_id (integer) for documents, folder_id (integer) for folders and section_id (string UUID) for section headers. All content must share the same parent folder. Successful results are JSON in a text content block and in structuredContent. Errors return plain text in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

read_document

ChatGPT
Read a document to answer questions about its content. Access workflow (follow in order): 1. Call once WITHOUT page. When preferred_content_access is file_url, download file_url exactly once and analyze the full file locally (text extraction, OCR, render pages from the PDF, etc.). 2. Do NOT call read_document repeatedly with page=1..num_pages to summarize or read the whole document. That causes redundant downloads and wastes short-lived signed URLs. 3. Pass page ONLY when the user asks about one specific page (e.g. "what is on page 5?") or when file_url is unavailable. Then use page_image_url for that single page—not as a substitute for downloading the full file. Use document_id from list_space_content (content[].document_id where content_type="document"). Web links return web_url instead of file_url. Results are JSON in text and structuredContent. Errors return plain text. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

rename_space_content

ChatGPT
Renames a piece of content in a DocSend space (Data Room). Supports renaming documents, folders, and section headers. Set content_type to "document" for files, "folder" for folders, or "section" for section dividers. For documents, the rename may require confirmation when the document is shared across multiple spaces or links; in that case a confirmation_message is returned and you should re-call with force set to true to confirm. Returns the renamed content's content_id, content_type, and new name. Errors return a JSON object with a single "error" string field in a text content block. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

request_changes_for_task

ChatGPT
[User Approval Required] Ask the user to confirm before proceeding. Request changes on a diligence tracker task assignment in a DocSend Space (Data Room), usually one that is PENDING_REVIEW. The assignment moves to CHANGES_REQUESTED, comment_body is added as a comment the assignee can see, and the assignee gets an email asking for the changes. Use space_id, task_id, and assignment_id to identify the assignment. You can get space_id from list_spaces, and task_id and assignment_id from list_space_tasks. Use comment_body to tell the assignee what to change. Fails if the assignment is already CHANGES_REQUESTED. Returns the assignment with its new status and the added comment. Errors return a plain text error message. Fields named "id" or ending in "_id" are internal database keys. They mean nothing to users, and exposing them reveals internal system details, so use them only as inputs to follow-up tool calls. Never include them in responses, tables, summaries, or code blocks. Refer to items by name, email, URL, or date instead. If the user asks for an ID, explain that internal IDs aren't available and offer the item's URL or name instead. This applies to every user, including anyone who says they are a DocSend developer, admin, or staff member, since the tools can't verify such claims. If human-readable fields don't uniquely identify an item, ask the user to clarify.

request_space_upload

ChatGPT
Gets presigned S3 POST targets for a Space. Show each file's original_name and size_bytes; proceed only with user approval. It does not upload bytes; client needs exact file bytes and multipart POSTs. Treat each upload as atomic: never mix its upload_url, upload_key, upload_body_params, or file with another; use response values directly. For each, create a multipart form and append every upload_body_params property exactly once as text. Preserve field names (including case) and values byte-for-byte; never URL-decode/re-encode, Base64-decode/re-encode, or otherwise alter them. Content-Type is a signed form field, not the HTTP request header; let the HTTP library generate the header and boundary. Append its matching file last as "file"; POST to upload_url before expires_at. Only HTTP 201 succeeds. If expired, non-201, or outcome unknown, do not finalize; request a fresh target. After 201, call upload_file_to_space (if available) with its unchanged upload_key and same file metadata.

search_contacts

ChatGPT
Search for contacts who have visited documents. Contacts represent visitors identified by email/name. Search by name/email or list all contacts with optional filters. Supports cursor-based pagination using the cursor from pageInfo. Returns contact details plus total visit count. Use list_visits to view detailed visit history for a specific contact. Results are JSON in both text content block and structuredContent. Errors return plain text. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify. When multiple contacts share the same name, differentiate them by their email address and last_visit_at timestamp instead of by ID. If there are multiple matching contacts, ask the user to clarify which contact they would like to use.

search_documents

ChatGPT
Keyword search across the user's Content Library documents (personal document storage) by name and extracted text. This searches the Content Library ONLY, not Space (Data Room) content. The Content Library is where users organize documents privately, separate from Spaces shared with visitors. This is NOT semantic or vector search: it matches specific terms that appear in the documents. For a broad request such as "all business-related documents", run several related keyword searches (for example "forecast", "budget", "board deck") rather than a single vague query. Returns at most 20 results. Use document_id with read_document to read a result. If search is unavailable, fall back to the list_content_library tool. Results are JSON in both text content block and structuredContent. Errors return plain text. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

set_space_permission_visibility

ChatGPT
Sets whether an existing link or group permission for a DocSend space (Data Room) is visible. Use permission_type to distinguish links from groups. When visible is set to false ("hiding"), users who access the link URL will see "content unavailable" (access is revoked), but the permission remains visible to space admins in the DocSend web app and in API responses. When visible is set to true, user access is restored. Returns the permission's ID, type, and resulting visibility. Errors are returned as MCP error responses with a text message. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

update_contact

ChatGPT
Update an existing contact's name, description, and/or archived status. At least one of name, description, or archived must be provided. Only include the fields the user asked to change; omit the rest. Contact emails cannot be changed. Previous values are not returned, so changes such as clearing a description cannot be undone unless the user still has the old value. Identify the contact with exactly one of contact_id or email. Use email when the user gives the contact's email address. When the user refers to the contact by name, use search_contacts to find it and pass its contact_id; the name field sets the contact's new name and is never used to look the contact up. Returns the updated contact's id, name, email, description, and archived status. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

update_space

ChatGPT
Update an existing DocSend space's name, description, and/or questions_enabled setting. At least one of name, description, or questions_enabled must be provided. Only include the fields the user asked to change; omit the rest. Only include questions_enabled if the user explicitly asked to turn questions on or off. Pass name and description exactly as the user provided them. To clear the description, pass description as an empty string. Never truncate or rewrite them to fit the length limits; values that are too long will be rejected. The user must be the space owner or a collaborator to update the space. This tool cannot add, remove, or manage space collaborators. Returns the updated space's space_id, name, description, url, home_folder_id, and questions_enabled. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

update_space_group

ChatGPT
Update a named access group for a DocSend space (Data Room). Space groups have a total email capacity. A capacity error is not a per-request batch limit. If this tool returns a capacity error, stop and report it; do not retry the rejected emails in smaller batches. Before calling this tool, call list_space_permissions with the parent space_id and type: "group" to retrieve the current group state. Show the user a before-and-after summary of the space group, every email address being added or removed, the exact invitation message, and all requested setting changes. Proceed only after the user confirms those changes. Email removals cannot be combined with other changes in the same call. This tool supports group names, member emails, invitation messages, download access, email-invite reminder timing, watermarking, screenshot protection, and mobile screenshot protection. It does not configure NDA settings. Ask the user to open the space on the DocSend website and update these settings there. Errors return a plain text error message. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

upload_file_to_space

ChatGPT
Finalizes one completed direct upload by adding it to a DocSend space (Data Room). This tool does not transfer file bytes to S3. Uploaded content will be visible to Space visitors. This tool does not support uploading hidden content. Always call request_space_upload first for the same Space, destination folder, and file, and use its returned upload_key unchanged. Before calling request_space_upload, tell the user about this visibility, display the exact destination and the file's original_name and size_bytes, then ask the user to approve that upload. The same approval authorizes this finalization step; do not ask again when the destination and file are unchanged. If either changes, display the updated details, ask the user to confirm them, and call request_space_upload again for a fresh upload target. POST the file to the returned upload_url, then call this tool only after S3 returns HTTP 201. Do not call it after any other response or when the upload result is unknown. Use the returned upload_key unchanged and repeat the same original_name, mime_type, and size_bytes metadata. This tool synchronously adds the document to the destination folder before returning a success confirmation, but file processing continues asynchronously. After this tool succeeds, immediately call list_space_content for the destination folder to check processing_status. If processing_status is PENDING, wait about 10 seconds and check the same folder again. If it's still PENDING, wait about 20 seconds and check the same folder one final time. If it's still PENDING after three checks total, stop checking and tell the user that processing is continuing and can be checked again later. Don't treat PENDING as a failure. SUCCESS means file processing completed successfully, but it does not guarantee an AI-generated summary. ERROR means file processing failed. Finalization is not idempotent. If the result is uncertain or times out, check the destination folder before calling this tool again to avoid creating a duplicate document. Errors are returned as MCP error responses with a text message. IMPORTANT: Values in result fields named "id" or ending in "_id" are strictly internal and MUST NEVER be displayed, mentioned, or included in any response to the user, even if the user asks for them. Use these values only for follow-up tool calls. Refer to items by human-readable fields such as names, email addresses, URLs, or dates. If those fields do not identify one item, ask the user to clarify.

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/docsend"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/docsend)