Home/ Guides/OpenClaw Slash Commands — Quick Reference Guide

OpenClaw Slash Commands — Quick Reference Guide

Control your OpenClaw agent from Telegram, Signal, or WhatsApp. Send these commands as messages to interact.

💡 Quick Tips

⚡

Essential Commands

/help
Show available commands and help information
/help
/status
Show current agent status including provider usage, quotas, and model info
/status
/stop
Stop the current agent operation immediately
/stop
/reset or /new
Start a fresh conversation session. Optionally specify a model.
/reset
/new claude
/restart
Restart the OpenClaw gateway service
/restart
🧠

Model & Thinking

/model alias: /models, /<alias>
Switch to a different AI model. Accepts alias, provider/model, or provider name (fuzzy match).
/model claude
/model openrouter/anthropic/claude-opus-4-6
/think aliases: /thinking, /t
Control how much reasoning the model shows. Options: off, minimal, low, medium, high, xhigh
/think high
/t medium
/reasoning alias: /reason
Control reasoning display. Options: on, off, or stream (show as it generates)
/reasoning stream
/reason off
🎛️

Session Control

/compact
Summarize and free up context space. Optionally add instructions for how to summarize.
/compact
/compact focus on recent decisions
/queue
Control message queue mode. Options: normal, debounce:2s, cap:25, drop:summarize
/queue debounce:2s
/queue cap:25
/verbose alias: /v
Control detailed output. Options: on, full, or off
/verbose on
/v full
/elevated alias: /elev
Elevate permissions for exec commands. Options: on, off, ask, or full
/elevated ask
/elev on
/exec
Configure exec behavior: host (sandbox/gateway/node), security (deny/allowlist/full), ask (off/on-miss/always)
/exec host=gateway security=allowlist
/exec ask=always
🔊

TTS / Voice

/tts
Control text-to-speech settings. Options: off, always, inbound, tagged, status, provider, limit, summary, audio
/tts always
/tts status
/tts summary
💻

Bash / Shell

/bash
Run shell commands on the host machine. Host-only — requires proper permissions.
/bash ls -la
/bash curl https://example.com
🛠️

Utility Commands

/whoami alias: /id
Show your sender ID — useful to verify you're allowlisted
/whoami
/allowlist
List, add, or remove allowlist entries. Controls who can access the agent.
/allowlist list
/allowlist add @username
/context
Explain current context usage. Use detail for per-file, per-tool, per-skill, and system prompt breakdown.
/context
/context detail
/context json
/usage
Control usage footer in responses. Options: off, tokens, full, or cost
/usage cost
/usage tokens
/subagents
Inspect, stop, log, or message running sub-agents. Options: list, stop, log, info, send
/subagents list
/subagents stop all
/approve
Resolve exec approval prompts. Options: allow-once, allow-always, or deny
/approve abc123 allow-once
/approve xyz789 allow-always
/skill
Run a skill by name. Optionally pass input to the skill.
/skill weather
/skill weather New York
/activation
Control when agent responds in groups. Options: mention (when @mentioned) or always
/activation mention
/activation always
/send
Control message sending. Options: on, off, inherit. Owner-only.
/send off
/config
Persist configuration to disk. Options: show, get, set, unset. Owner-only, requires commands.config: true
/config show
/config set theme dark
/debug
Show or set runtime overrides. Options: show, set, unset, reset. Owner-only, requires commands.debug: true
/debug show
/debug set verbose true

🔧 Troubleshooting

Agent stopped responding Try /stop then resend your message
Agent seems confused Use /reset to start a fresh session
Wrong model selected Use /model <name> to switch models
Context too full / hitting limits Use /compact to summarize and free space
Agent running slow Check /status for rate limits or cooldowns
Messages out of order Use /queue to check and adjust queue mode
Can't run commands Check if you're allowlisted with /whoami to verify your sender ID
Gateway down / not responding Run openclaw gateway restart from your terminal

Next steps

Keep building your setup with these guides and references:

Stay in the loop

The 5-minute weekly read that makes your OpenClaw smarter.