Chat Commands

Chat commands are slash commands you can use in any channel (WebChat, Telegram, Slack, Teams, WhatsApp, TUI) to control Octipus. Commands execute immediately without an LLM call.

All Channels

Command	Description
/help	Show available commands
/status	Check session status, running agents
/expert [name]	List experts or switch to a specific expert
/stop	Stop all running agents in the session
/clear	Clear conversation context and start fresh
/models	List available models
/plan	Start an interactive project planning questionnaire
/cost	Show token usage and cost for the session
/cancel	Abort a multi-step command (e.g., active /plan)

TUI / Gateway Commands

Command	Description
/cost	Show cumulative token usage and cost
/diff	Show git diff --stat for workspace changes
/version	Show assistant and Bun version
/compact	Compact session context (summarize old messages)
/exit	Exit the TUI

Channel-Specific

Command	Where	Description
/start	Telegram	Initialize bot connection
/link	Telegram, Slack, Teams, WhatsApp	Get a 6-character code to link channel account to web account

Expert Switching

Switch between pre-built expert personas from any channel:

/expert                    → List all available experts
/expert Technical Writer   → Switch to Technical Writer
/expert reset              → Return to auto-routing

When an expert is active, all messages bypass the orchestrator’s classifier and go directly to that expert’s agent with its specialized prompt, domain knowledge, and rules. Expert selection persists across messages in the session.

Permission Prompts

When an agent needs to use a tool that requires approval, you’ll see a permission prompt:

Telegram/Slack/WhatsApp:

Permission required: the agent wants to use "shell".
Command: docker compose up -d
Reply "yes" to allow or "no" to deny.

TUI:

Permission: shell → docker compose up -d  (y/n)

Channel Feedback (Emoji Reactions)

On external channels (Telegram, Slack, WhatsApp), Octipus provides real-time emoji reactions:

Phase	Meaning
Received	Message received
Working	Role-specific indicator (coding, research, etc.)
Tool use	Tool-specific indicator during execution
Permission	Waiting for user approval
Done	Success or error

A typing indicator is also sent while the agent works.