Your first chat

Ask something

Type a message and press Enter. A few ideas to start with:

• "What's on my calendar tomorrow?"
• "Summarize the unread e-mails from this week and draft replies to the two most urgent."
• "Which of our automations failed yesterday, and why?"
• "Research the latest pricing changes of our three main suppliers and make a one-page report."

The reply streams in. Each tool the agent uses shows as a row that turns from running to done or failed, with the automation runs and app actions it made underneath.

Approve what it proposes

When the agent wants to change something (send an e-mail, update a record, run an automation, create a schedule) it shows an approval card with exactly what will happen. Approve it, cancel it, or answer in words: a short "yes" runs what was proposed.

Open a folder

To work on code, open a folder: the folder button in the app, /workspace ~/code/my-project, or run crewdesk inside the folder. The agent can then read, edit, run and use git there, under your permission mode. If the folder has an AGENTS.md (or CLAUDE.md), the agent follows it; /init asks the agent to write one.

Useful commands

Command	What it does
/new	Start a new chat
/plan	The next message gets a plan to approve before anything changes
/model strong	Use the strongest model tier for this chat
/assistant NAME	Start a chat with one of your assistants
/activity	What Crewdesk did for you
/stop	Stop the running reply
/help	All commands, and your organization's skills

The full list is in Slash commands.