Data sent to AI

What goes where

• Your workspace stays on this computer. It is one file in your Contexl folder. Core never sends it anywhere, and Core itself never calls an AI model or the internet.
• What Core shows Claude goes to Anthropic. When Claude runs a Core command in Claude Code, the command's output (briefs, signals, parts of your strategy, digests) becomes part of your conversation and is sent to Anthropic under your own Claude plan and its data terms. Check your plan's privacy settings.
• Accounts and contacts go to Anthropic only when you work on them. When you ask Claude about companies or people (fit, contacts, account reports), the names and details it shows become part of your conversation, under your Claude plan. Other commands, such as the weekly competitor analysis, the message map and the report, show no contact details. You are responsible for the personal data you put into Contexl (under GDPR, you are its controller).
• Call notes and transcripts. When you add a call, Claude reads it, so it goes to Anthropic under your plan. Contexl keeps a short summary and the quotes you approve as signals. A raw transcript stays in the workspace for 90 days from when you added it; Core then removes it at the start of a session. A backup made before that still holds it until the backup is replaced.
• Output Skills ask before they contact other services. A Skill that collects information (for example web search, or a connector) tells you which service gets which queries, and waits for your approval.
• Every AI routine is recorded. Core keeps, per run, the routine, the AI provider and model, the prompt version and a summary of what was sent. The "Data sent to AI" page lists them.
• Backups go where you choose. A backup is a copy of the workspace file in a folder you pick. If that folder is synced (Dropbox, OneDrive, iCloud), the copy goes to that service too.