Enterprise controls

Enterprise plans add the controls security and procurement teams ask for: where AI processing happens, and how long workspace data lives.

Private LLM (AWS Bedrock)

By default the assistant uses Anthropic's Claude models via Anthropic's API. Enterprise workspaces can instead be routed through a private deployment on Amazon Bedrock: the same Claude models, served inside AWS rather than through the public API, with AWS's data handling (no training on your content, no retention beyond serving the request in standard configurations). Everything else — permissions, approvals, audit, billing — is unchanged; if a specific model isn't available in the chosen region, the assistant falls back gracefully to the nearest available Claude model rather than failing.

Data region selection

Bedrock-routed workspaces choose the AWS region their AI processing runs in, which is how data-residency requirements are met — for example keeping model processing in a specific jurisdiction. Model availability varies by region; we'll confirm the best fit for your residency requirements during onboarding.

Data retention controls

Set a rolling retention window (for example 30, 90 or 365 days) and workspace content older than the window is permanently deleted on a daily schedule: chat sessions and their messages, stored files, and external-AI (MCP) activity records. What's deliberately kept regardless: the audit log (who did what, when), usage records and billing history — the record of actions remains while the content itself expires. Without a window configured, data is retained while the account is active, and deleted 30 days after cancellation as standard.