Skip to main content
Feb 13, 2026

Features

Syncs
  • Progress tracking: Data syncs now provide real-time progress updates during execution.
  • Cancel sync: Running syncs can be cancelled from the UI via a dedicated button.
Data sources
  • Credential name: Newly created credentials can now be named directly during the “new data source” flow.
Retrieval
  • Include sources: Retrieval API and MCP can now return where context was derived from. See Improvements and API changes for configuration details.
Syncs
  • Durability: Made syncs more durable and reliable.
  • Automatic initial sync: Syncs now start automatically when a data source is added.
  • Automatic sync after edit: Syncs now start automatically when filters are updated on a data source.
  • Sync already running: When a sync is triggered for a source that is already syncing, the current sync can be stopped or the new one skipped.
MCP
  • Improved vault selection: The “select vault” tool reply is now more informative for agents and works correctly with ChatGPT.
  • Multiple clients: Added support for connecting multiple MCP clients of the same type (e.g. Cursor, Claude) to different vaults.
  • Expired sessions: MCP sessions expire 7 days after creation and are now hidden from the UI once expired.
  • Source information: Added includeSources parameter to the retrieval tool so the LLM can request source information in the response.
API keys
  • Expiry column: Added an expiry column to the API keys table.
  • Revoked keys: Revoked API keys are now hidden from the API keys table.
  • Expired keys: Expired API keys can easily be deleted.
  • Retrieval logs: Retrieval logs now display the name of revoked or deleted API keys instead of “Deleted API Key”.
  • Automated syncs: Fixed an issue where they could fire twice (with one cancelled); exactly one sync runs now.
  • Website data sources: Automated syncing for websites connected via API now works.
  • Delete vault: Fixed a bug in the vault deletion. Button now works as expected.
  • API key instructions: Fixed creating and updating instructions for API keys.
  • knowledgeGraphIdvaultId: Renamed in ingestion and retrieval endpoints (ingestion: unstructured, website, any; retrieval).
  • workspaceId: Removed from ingestion and retrieval; use vault-scoped paths (ingestion: unstructured, website, any; retrieval).
  • includeSources: Added to retrieval (default false). When true, the response includes where context was derived from. Breaking change on Feb 27: default will become true. Set includeSources = false in requests or adopt source information before then.