When Claude is active

Claude works in Turndesk only when you bring it in, sees only what you allow, and changes nothing in Easy8 or Redmine without your click.

Does Claude run on its own?

Never.

A model works only in these three moments. Without your own key or a connected agent, none of them can happen.

  • You press Enter in the Ask Claude panelWith your own Anthropic or OpenAI key.

  • You paste a question into the Claude appThe Ask in Claude button copies it; Turndesk sends nothing.

  • An agent you started calls TurndeskClaude Code, Codex or another MCP client.

  • Never at startup, on a timer, on a notification or in Turndesk Tray.

What does the model get?

Person-2, not Kai.

At the default access level, names become Person-N before the question reaches the model. Passwords, keys, phone numbers, IBANs and card numbers are removed at every setting.

A comment on #4821

Kai MorenoPerson-2 asked Lena ParkPerson-3 to review the refund fix before Friday. LenaPerson-3 can test it with card 4111 1111 1111 1111[card].

Can Claude change Easy8 or Redmine?

Only with your click.

Claude leaves a proposal with its reasoning. Time, estimates, comments and new tickets reach Easy8 or Redmine when you accept them.

A proposal to log 2.5 hours on #4821, with the agent's reasoning and Log and Dismiss buttons

What does Claude remember?

What you can see.

Up to 30 short lines about you, never about anyone else. Each change shows under the answer with Keep and Discard, and nothing is saved until you choose Keep.

Ask Claude: a short answer, then Remembered and Updated lines, each with Keep and Discard

How do you turn it off?

Three switches.

In Settings, AI agents: remove the key or disconnect the agent, switch off chat memory, or choose Details only to keep comments away from the model. The full detail is in the privacy policy.

Agent data access: Everything, No people, Details onlyChat memory in Settings: Remember between chats, three remembered lines with Edit and Forget, and Clear