Overview
Bring AI tools to the Trust Center. Search and manage Knowledge Base entries, approve access requests, and pull approved Trust Library content into your AI workflows from Claude, Cursor, ChatGPT, or any MCP-compatible client.
What is it?
An MCP (Model Context Protocol) server that connects SafeBase to AI clients
The MCP server lets you use natural language to manage your Trust Library content without leaving your AI tool of choice.
Works with any MCP-compatible client
Who can use it?
You must be an active organization member in SafeBase
Disclaimer
Organizations can enable the MCP server with read-only or read + write permissions.
With read + write enabled, you accept the risk of unintended changes to your Trust Library content. Some tools overwrite or remove data:
questionnaires_retry_autofill_rowsqueues eligible rows for a new AI answer, replacing what is there now,questionnaires_edit_answer_and_commentclears a row's AI answer sources and confidence when you change it,tags_deleteandtranslations_applyin delete mode remove data, and the Trust Center content tools publish live with no draft step.
Setup
Prerequisites
An admin must enable either read-only or read + write in Settings > Generative AI > MCP for your organization.
SafeBase MCP server URL: https://app.safebase.io/api/mcp
Client-specific setup
Claude (claude.ai) and Claude Desktop
Team/Enterprise: An owner or admin must first add the connector via Organization settings → Connectors → Add → Custom → Web. Once added, members can connect from the steps below.
Open Customize → Connectors
Click the + button → Add custom connector
Enter the SafeBase server URL above
Sign in to SafeBase when prompted
To use it in a chat: click the + button (lower left) → Connectors → enable SafeBase
ChatGPT (Plus / Pro / Business / Enterprise / Edu)
Custom MCP connectors require Developer Mode. For Team/Enterprise/Edu, a workspace admin must first enable it under Admin Workspace Settings → Permissions & Roles → Connected Data → Developer mode.
Open Settings → Connectors
Scroll to Advanced and toggle Developer mode on
Click Create
Enter the SafeBase server URL above
Sign in to SafeBase when prompted
Claude Code
Run:
claude mcp add --transport http safebase <https://app.safebase.io/api/mcp>
Then inside Claude Code, run /mcp to complete the browser sign-in.
Cursor
Add the SafeBase server to ~/.cursor/mcp.json (or .cursor/mcp.json in your project):
{ "mcpServers": { "safebase": { "url": "<https://app.safebase.io/api/mcp>" } } }Restart Cursor and sign in to SafeBase when prompted.
Tools Supported
Ask AI
ai_ask— Ask a natural-language question grounded in your approved Trust Library content
Accounts
accounts_list— List accountsaccounts_get— Get an account by IDaccounts_create— Create an accountaccounts_edit— Edit an account
Account Members
accounts_find_member— Search account members across accountsaccounts_add_members— Add members to an account
Trust Center access requests
access_requests_list— List access requestsaccess_requests_approve— Approve an access request
Trust Center Updates
tcu_list_topics— List update topicstcu_get_topic— Get a topic by IDtcu_create_topic— Create a new topictcu_update_topic— Update a topictcu_create_update— Post an update under a topic
Trust Center content
trust_center_content_get— Retrieve Trust Center products and itemstrust_center_content_update_item— Update a Trust Center item's title, description, maturity, or shared products. Changes go live immediately.trust_center_content_link_documents— Link or unlink documents on a Trust Center item. Changes go live immediately.
Content Collections
content_collections_list— List content collectionscontent_collections_get— Retrieve one content collection by ID, including its sources and tag filterscontent_collections_create— Create a content collectioncontent_collections_edit_tags— Add or remove tag filters on a content collection
Documents
documents_download— Retrieve short-lived download links for one or more documentsdocuments_list— List documentsdocuments_get— Get a document by IDdocuments_create— Create a documentdocuments_update— Update a documentdocuments_upload_request— Create a document upload request
Knowledge Base
kb_search— Search KB entrieskb_list— List KB entries with filters (tag, source, product, assignee, health, review cadence)kb_create_entry— Create a new KB entrykb_update_entry— Update an existing KB entry's contentkb_get_suggestions— Pull AI-generated KB suggestionskb_handle_suggestion— Approve or dismiss a KB suggestionkb_sources_list— List available KB content sourceskb_get_entries_assignees— Get assignees for KB entrieskb_set_assignees— Set assignees on a KB entrykb_create_suggestion— Create a KB suggestion for human review instead of writing to the KB directlykb_bulk_edit_tags— Add or remove tags on multiple KB entries at once, leaving other tags in place
Questionnaires
questionnaires_list— List questionnairesquestionnaires_get— Get a questionnaire by IDquestionnaires_get_sheets— List the sheets in a questionnairequestionnaires_get_sheet_rows— Get the rows in a sheetquestionnaires_get_row_assignees— Get assignees on a rowquestionnaires_set_row_assignees— Set assignees on a rowquestionnaires_bulk_assign_rows— Bulk-assign rows to membersquestionnaires_edit_answer_and_comment— Update the answer and/or comment on a questionnaire row. This clears the row's AI answer sources and confidence, the same as editing it in the app.questionnaires_retry_autofill_rows¹ — Rerun AI autofill on selected rows. This overwrites current answers and comments.
Org / admin
org_get_me— Get your own profileorg_get_my_permissions— Get your effective permissionsorg_get_info— Get org infoorg_get_settings— Get org settingsorg_get_nda— Get the NDAorg_list_members— List org membersorg_invite_members— Invite new membersorg_list_invites— List pending invitesorg_list_roles— List rolesorg_get_role— Get a role by IDorg_create_role— Create a roleorg_update_role— Update a role
Trust Library reference
tags_list— List available tagsproducts_list— List available productssearch_assignees— Search assignable members and groups by name or email.tags_create— Creates a tag, or returns the existing tag if one with that name already existstags_update— Renames a tagtags_delete— Deletes one or more tags. Tags still linked to a content collection are skipped.
Trust Library Search
trust_library_search— Search Trust Library content by relevance and return matching items without generating an answer
Permission Profiles
permission_profiles_link_documents² — Link documents to a permission profile
Translations
translations_explore³ — Review Trust Center translations, including supported languages, coverage, and existing translated values.translations_apply³ — Add, update, or remove translated values for Trust Center content
¹ Only available to orgs with Questionnaire Assistance.
² Behind the enableDocumentInPermissionProfile feature flag.
³ Behind the enableClientTranslationsExtApi feature flag.
Reviewing a vendor’s Trust Center? Use the Vendor Trust Center MCP User Guide.
