Slash commands

How it works

Type / at the start of a message and a menu of available commands appears. Commands that need arguments open a small form, so you do not have to memorize the argument order; they are submitted like messages and their result lands in the thread. /creative works differently: it opens a Creative picker and inserts your selection as a link in the draft, which you can continue editing before you send the message yourself.

Built-in commands

/calendar (also /cal) creates a local calendar event from a date and a memo, accepting relative dates like tomorrow, +3days, or +mon. When Google Calendar is connected, Collavre also attempts to sync the event there. /topic creates a named topic and can assign a primary agent to it. If that name already exists, it targets the existing topic instead: mentioning an agent assigns or replaces its primary agent, while omitting the mention releases any existing primary agent. Assigning, replacing, or releasing a primary agent requires write permission. /work walks the child Creatives depth-first and has an agent execute a task on each eligible one, skipping Creatives that are already complete or have an active task. /compress summarizes the current topic's messages and removes the originals. /creative inserts a link to another Creative, which in an AI chat also injects that Creative and descendants up to the agent's configured child-depth limit as context.

MCP tools

MCP tools registered in your workspace show up in the same menu below the built-in commands and run through the same command form. The tools you see depend on which ones are registered and which ones you have permission to use.

Tips

• Whatever you had typed is kept. A form-backed command temporarily stashes the draft, submits the command alone, and restores the draft after the send settles. Commands without a form stay in the message box for you to edit and send.
• /compress requires a topic, write permission, and an available AI agent. Collavre tries the topic's primary agent, the policy primary agent, then any AI agent shared on the Creative with comment permission.
• Commands are submitted as messages, so they generally need comment permission, and they are unavailable on an archived Creative. Assigning, replacing, or releasing a /topic primary agent additionally requires write permission.