When chatting with an OpenClaw Agent, use / commands to control its behavior.
Model Switching
| Command | Description | Example |
|---|
/model <alias or ID> | Switch the model for the current session | /model opus |
/model list | List all available models and aliases | /model list |
/model reset | Reset to the default model | /model reset |
Common aliases (configured in models.aliases):
/model opus # claude-opus-4-6 (strongest reasoning)
/model sonnet # claude-sonnet-4-5 (balanced)
/model haiku # claude-haiku (fastest)
/model gpt # gpt-5.2
/model ds # deepseek-chat
Context Management
| Command | Description |
|---|
/compact | Compress the current context (keep summary, save tokens) |
/compact <instruction> | Compress with a custom instruction, e.g. /compact keep all code snippets |
/clear | Clear the current session context (use with caution) |
/context | Show current context usage (token count) |
/memory | View the agent's persistent memory |
/memory add <content> | Manually write an entry to agent memory |
/memory clear | Clear all persistent agent memory |
Thinking & Reasoning
| Command | Description |
|---|
/thinking on | Enable extended thinking mode (Opus 4.6+ only) |
/thinking off | Disable extended thinking mode |
/thinking budget <n> | Set max thinking token budget, e.g. /thinking budget 5000 |
Note: /thinking only works on models that support extended thinking (e.g. Claude Opus 4.6).
| Command | Description |
|---|
/activation | View currently active tools / Skills |
/activation <skill> | Activate a specific Skill |
/tools list | List all available tools |
/tools enable <tool> | Enable a specific tool |
/tools disable <tool> | Disable a tool (for this session only) |
System Control
| Command | Description |
|---|
/status | Show agent status (model, context, tools) |
/restart | Restart the current agent session |
/debug | Toggle debug mode for verbose logging |
/help | Show all available slash commands |
/version | Show current OpenClaw version |
Files & Code
| Command | Description | Example |
|---|
/read <path> | Read a file and send its contents to the agent | /read ./config.json |
/run <command> | Execute a shell command and return the result | /run ls -la |
/paste | Enter paste mode for multi-line input | /paste |
Usage Examples
Quick model comparison
/model haiku # fast draft
(send your task)
/model opus # deep processing
(send complex task)
Compress long conversations
/context # check current token usage
/compact keep all completed task lists and code blocks
(continue the conversation without hitting context limits)
Activate a Skill for a specific task
/activation bird # activate the Twitter bird skill
Find the most popular AI tweets from today
Enable deep reasoning
/model opus
/thinking on
/thinking budget 8000
Analyze the performance bottlenecks in this code...
Configuring Aliases (in openclaw.json)
{
agents: {
defaults: {
models: {
"anthropic/claude-opus-4-6": { alias: "opus" },
"anthropic/claude-sonnet-4-5": { alias: "sonnet" },
"anthropic/claude-haiku-4-5": { alias: "haiku" },
"openai/gpt-5.2": { alias: "gpt" },
"deepseek/deepseek-chat": { alias: "ds" },
},
},
},
}
Tip: Slash commands work across all channels — Telegram, Feishu, Discord, and more.