Chapter 11: All Chat Commands

OpenClaw users interact with the gateway through chat commands — special messages that start with / and control the session, agent behavior, and gateway status. These commands work across every connected channel: WhatsApp, Telegram, Slack, Discord, Teams, and all others.

This chapter documents every built-in chat command with full examples.

How Chat Commands Work

When a user sends a message starting with /, the OpenClaw gateway intercepts it before passing it to the AI agent. The gateway processes the command directly and returns a system response. Commands are not sent to the AI model — they control the framework around the model.

User sends: /reset
  ↓
Gateway intercepts (does NOT call AI)
  ↓
Gateway clears session for this user
  ↓
Gateway replies: "Session cleared. Starting fresh."

The Complete Command Reference

`/status`

Purpose: Check the gateway's current status — which channels are online, how many sessions are active, and what model is assigned to your workspace.

Usage:

/status

Response:

OpenClaw Gateway Status
━━━━━━━━━━━━━━━━━━━━━━
Version: 1.4.2
Uptime: 3h 47m

Channels:
  ✅ telegram — connected
  ✅ slack — connected
  ❌ whatsapp — disconnected (reconnecting...)

Your Session:
  Workspace: dev-team
  Agent: claude-sonnet
  Messages in session: 12
  Session age: 23 minutes

Tip: Use /status first when something feels wrong — disconnected channels and wrong agents are the most common issues.

`/new`

Purpose: Start a completely new conversation session. Clears all previous context. The agent will have no memory of your previous messages in this session.

Usage:

/new

Response:

New session started. All previous context cleared.
How can I help you?

When to use: When switching to a completely different task and you do not want previous context to influence the agent's answers.

Alias: /reset does the same thing.

`/reset`

Purpose: Identical to /new — resets the current session, clearing all conversation history.

Usage:

/reset

Response:

Session reset. Starting fresh.

`/compact`

Purpose: Compresses the current session's conversation history. Instead of sending all previous messages to the AI (which costs tokens), the gateway summarizes the conversation so far and replaces the full history with a condensed version.

Usage:

/compact

Response:

Session compacted.
Conversation summarized from 47 messages → 1 summary.
Token usage reduced by ~73%.

Summary: "Working on a React authentication system. Implemented JWT login flow. Currently debugging token refresh logic in AuthService.ts."

When to use:

When your session is getting very long and responses are slowing down
When you have finished one task and want to preserve context but reduce cost before moving to the next

`/think`

Purpose: Switches the agent to extended thinking mode. In this mode, the AI takes more time before responding, showing its reasoning process. Produces better results for complex problems at the cost of higher token usage.

Usage:

/think

Or combine with a question:

/think what is the most efficient algorithm for finding duplicates in a list of 10 million items?

Response:

[Thinking mode enabled]

Let me reason through this step by step...
[Extended reasoning shown]

Answer: For 10 million items, a hash set approach gives O(n) time and O(n) space...

Disable thinking mode:

/think off

`/verbose`

Purpose: Enables verbose mode — the agent shows more details in its responses, including intermediate steps, tool calls made, and reasoning explanations.

Usage:

/verbose

Example response in verbose mode:

[Verbose mode ON]

You asked: "List all TypeScript files modified this week"

I'll run a bash command to find recently modified files.

Running: find . -name "*.ts" -newer $(date -d "7 days ago" +%Y-%m-%d) -type f
Output: [list of files...]

Result: Found 14 TypeScript files modified in the last 7 days.
[List follows...]

Disable:

/verbose off

`/trace`

Purpose: Enables full trace logging for your session. Every API call, tool execution, token count, and latency measurement is logged. Primarily used for debugging.

Usage:

/trace

What it enables:

Logs every message sent to and received from the AI model API
Shows exact tool call inputs and outputs
Shows token usage per message
Shows latency for each step

Disable:

/trace off

Note: Trace output is in the gateway's log file (~/.openclaw/logs/openclaw.log), not in the chat. This command is for developers debugging agent behavior — not for regular users.

`/usage`

Purpose: Shows your token usage and estimated API cost for the current session.

Usage:

/usage

Response:

Session Usage
━━━━━━━━━━━━
Messages: 23
Input tokens: 18,432
Output tokens: 4,891
Total tokens: 23,323

Estimated cost (claude-sonnet):
  Input: $0.0553
  Output: $0.0734
  Total: $0.1287

Session started: 2h 14m ago

Note: Costs are estimates based on current Anthropic pricing. Actual billing is from your API provider.

`/restart`

Purpose: Restarts the OpenClaw gateway daemon. Use when configuration changes have been made or when channels are behaving unexpectedly.

Usage:

/restart

Response:

Restarting gateway...
[Gateway goes offline briefly]
Gateway restarted. All channels reconnecting.

Important: This restarts the entire gateway, disconnecting all active sessions briefly. All users will see a brief outage. Use during low-traffic periods.

Permission: By default, only users in workspaces with adminCommands: true can run /restart.

Admin-Only Commands

Some commands are restricted to admin users. Configure admin access in openclaw.json:

{
  "workspaces": [
    {
      "id": "admin",
      "adminCommands": true,
      "allowlist": ["987654321"]
    }
  ]
}

Admin commands:

Command	Description
`/restart`	Restart the gateway daemon
`/reload`	Reload config without full restart
`/ban [user-id]`	Remove a user from all allowlists
`/sessions`	List all active sessions across all workspaces
`/logs [n]`	Show last n lines of the gateway log

Workspace-Specific Commands

Some commands behave differently per workspace:

{
  "workspaces": [
    {
      "id": "dev",
      "disabledCommands": ["/restart", "/trace"],
      "enabledCommands": ["/status", "/new", "/reset", "/compact", "/usage"]
    }
  ]
}

Use disabledCommands to prevent regular users from running sensitive commands.

Command Reference Table

Command	Category	Description	Token Cost
`/status`	System	Show gateway and session status	None
`/new`	Session	Start a new session	None
`/reset`	Session	Reset current session	None
`/compact`	Session	Compress session history	Small (summarization)
`/think`	Agent	Enable extended reasoning	High
`/think off`	Agent	Disable extended reasoning	None
`/verbose`	Agent	Show detailed step-by-step output	Low
`/verbose off`	Agent	Disable verbose output	None
`/trace`	Debug	Enable full trace logging	None
`/trace off`	Debug	Disable trace logging	None
`/usage`	Billing	Show token and cost usage	None
`/restart`	Admin	Restart the gateway	None
`/reload`	Admin	Reload config without restart	None
`/sessions`	Admin	List all active sessions	None
`/logs [n]`	Admin	Show last n log lines	None

Channel-Specific Notes

WhatsApp: Commands work exactly the same. Type /status and send it just like any other message.

Telegram: Commands appear in Telegram's built-in command autocomplete if you configure them via BotFather (/setcommands).

Slack: Commands appear as slash commands. The first time, Slack may ask you to approve the /status, /new, etc. commands for your workspace.

Discord: Native slash commands are registered via openclaw discord register-commands. They appear in Discord's autocomplete menu.

Next: Chapter 12 — Configuration File Deep Dive — Every field in openclaw.json explained with real-world examples.