September 21, 2026

42 changes across 6 repo s

New

  • Research content is now access-controlled per user and organization #627 Research documents, search results, and related endpoints are now served only to users and organizations with an active research grant. Users without access see research omitted from global search results rather than an error, and a new view_ui_research field on the user profile lets client applications gate the research UI accordingly.
  • AI-ready feed is now an independent toggle #839, #368 The AI-ready feed can be enabled or disabled on its own, without any dependency on the bulk research feed. Organization admins can manage each feed separately, and both feeds are reflected as distinct rows in the provider detail view.
  • Provider content search endpoint #859 Content providers can now search and filter their own document catalog via a paginated endpoint. Supported filters include processing status, publication date range, file received date range, filename, and external document ID.
  • Provider file export #854 Content providers can now export a CSV of all files in their registry, including filenames, received and updated dates, and processing status. The export runs asynchronously and delivers a download link via the existing export notification flow.
  • MCP users see an upgrade prompt in AI chat and authoring surfaces #675 Users on an MCP seat now encounter a read-only state with an “Upgrade to a full seat” overlay in the AI chat, prompt authoring, and live transcript areas. Audio playback and live transcript streaming are also gated, with the overlay linking to a contact-sales dialog.
  • Company Documents page now includes Research and Company Docs filter options #740 The document-type filter on the Company view’s Documents page has been expanded to include Research (entitlement-gated, matching the global Document Search) and Company Docs. Both options respect the same access rules applied elsewhere in the product.
  • Vendor integration management for organizations #620 Organization administrators can now browse available vendor integrations and enable or disable them for their organization through a self-service API surface. Platform administrators have additional endpoints to register and configure vendors globally.

Improvements

  • Token usage table now shows a cleaner, more relevant user list by default #638 The organization token usage table defaults to showing active members plus any user who consumed tokens in the selected period, hiding deactivated members with no usage. New include_inactive and has_usage query parameters let you adjust this view. The table’s totals always reconcile exactly with the organization summary.
  • Historical token pool allocations now reflect actual seat counts at the time #641 Past months’ pool limits, utilization percentages, and statuses are now calculated using the seat count that applied during each period rather than today’s count. This corrects cases where growing organizations saw historical allocations inflated by later seat additions.
  • Token pool reset time is now included in the API response #638 The organization token summary response now includes a pool_resets_at timestamp indicating when the current pool period ends, making it easier to display accurate reset timing in client applications.
  • Text selection on mobile transcript works as expected with native controls #741 On touch and pen devices, selecting text in an event transcript no longer triggers the highlight toolbar prematurely. The native OS selection handles now have room to operate, and tapping an existing highlight still opens the recolor and remove options.
  • Client detail panel closes automatically after a successful delete #373 Deleting a client from the detail panel now reliably closes the panel and keeps it closed. Switching directly between two open clients also no longer briefly shows the previous client’s data, and the panel now displays loading and error states.
  • REST API documentation no longer references research endpoints for general access #67 The REST API getting-started page and product card have been updated to reflect that analyst research is now an entitlement-gated feature rather than a general offering. Chat and MCP documentation, where research remains available to entitled users, are unchanged.
  • MCP tool description clarifies that a company scope is required for document searches #128 The find_company_docs tool now clearly states that a company identifier is required, removing guidance that suggested leaving it empty to search across all companies.

Fixes

  • Transcript PDF downloads now complete successfully #647, #648, #649 Downloading a transcript as a PDF now returns the file as expected across all environments.
  • Email verification now redirects to the dashboard as expected #860 Users who click an email verification link are now taken directly to the dashboard after confirming their address, rather than being sent back to the login page.
  • Export notification in the dashboard widget now renders as a link #375 The download link in the notifications summary panel on the dashboard now appears as a clickable action rather than displaying raw HTML text.
  • Provider client Slack notifications now show the correct organization name per client #863 When multiple clients are added at once, each client in the confirmation notification is now matched to its own organization name rather than potentially showing a mismatched name.
  • Public API responses now return the external client identifier #861, #867, #878, #877 API responses that include client identifiers now surface the external CRM client ID rather than an internal identifier.
  • Notification preferences load reliably even when stored data has an unexpected shape #643 Notification preference records with a malformed but recognizable structure are now handled gracefully on read, so the preferences surface loads consistently.

← All updates