What Are Slash Commands?

Every SuperClaws agent running on Telegram understands slash commands — short instructions you type that start with /. They are processed instantly, before the AI model even sees your message. Think of them as the remote control for your agent: session management, model switching, status checks, and more — all without crafting a prompt.

You are probably already using a few without realizing it. But most people only scratch the surface.

Your Top 5 Commands

1. /new — The Clean Slate

This is the most important command most users underuse. /new clears your entire conversation context and starts a fresh session. Your agent forgets everything from the current chat — old instructions, irrelevant background, accumulated noise — and starts clean.

Why this matters: context windows fill up. When they do, your agent gets slower, more expensive, and sometimes confused by old details that no longer apply. /new is the reset button.

Bonus: you can pass a model hint: /new grok-4.1-fast starts a fresh session on that specific model. Two birds, one stone.

2. /model — Switch Models Without Leaving the Chat

This deserves its own highlight because it is incredibly powerful. /model lets you switch between AI models mid-conversation without any config changes.

Send /model alone and you get a numbered picker showing available models. Send /model 3 to pick the third one. Or type a model directly: /model openrouter/x-ai/grok-4.1-fast.

Why this matters: different models have different strengths. Maybe you want a fast, cheap model for quick questions and a powerful one for complex analysis. /model lets you switch on the fly. The change persists for the rest of your session.

You can also use model aliases if you have configured them. For example, if you set an alias like fast for Grok, just type /fast and you are there.

3. /stop — Abort a Run

Sent a message you regret? Agent going down a rabbit hole? /stop immediately aborts the current run. No waiting, no hoping it figures itself out. Just /stop and you are back in control.

4. /btw — Ask a Side Question

This one is underrated. /btw lets you ask a question about the current session without changing its context. It runs as a separate, ephemeral call — your agent answers it, but the answer is not written to conversation history and does not affect future responses.

Example: /btw what are we working on right now?

Use it when you want a quick clarification without derailing the main task.

5. /status — The Health Check

Send /status and your agent replies with a snapshot: current model, session age, token usage, cache percentage, and provider quota if available. It is the fastest way to understand what is happening under the hood when your agent starts acting sluggish.

Two More You Should Know

/restart — Reboot the Agent

More powerful than /new. /restart restarts the entire OpenClaw gateway — the agent process itself, not just the session. Use this when something feels off at a deeper level: tools not responding, plugin errors, or the agent is in a weird state. Enabled by default.

/help and /commands — The Full Menu

Want to see everything available? Send /help or /commands. This gives you the complete list of commands your agent supports, including any skill-based commands. It is always up to date because it pulls from the live runtime — not a static docs page.

This is also how you discover commands you did not know existed. /tools shows what tools your agent can use right now. /think high cranks up reasoning depth. /verbose on shows extra debugging info.

Start with /commands. You will be surprised what is already there.

Quick Reference

CommandWhat it does
/newClear context, start fresh session
/new model-nameNew session on a specific model
/modelShow model picker
/model 3Switch to model #3 from the list
/stopAbort the current run
/btw questionAsk a side question without affecting context
/statusShow session health and usage
/restartRestart the OpenClaw gateway
/helpShow all available commands
/commandsSame as /help — full command list
/think highIncrease reasoning depth
/verbose onEnable detailed output for debugging
/usage tokensShow per-response token usage

The Pattern

Here is how experienced SuperClaws users actually work:

  1. Start with /model to pick the right model for the task
  2. Do your work in a session
  3. Use /btw when you need a quick side answer
  4. Use /stop when things go sideways
  5. Hit /new when starting a different topic
  6. Check /status when performance feels off
  7. Hit /help when you forget what else is available

These commands turn your Telegram chat from a passive messaging interface into a fully controllable agent console. Learn these five and you will move faster than 90% of users who never go beyond plain messages.