Edited messages are invisible
If you edit a Telegram message after sending it, the bot usually will not see the updated version. Send a new message instead.
Telegram Commands Cheat Sheet
30+ commands for daily use with your OpenClaw agent.
Essential Commands
| Command | What It Does | When To Use |
|---|---|---|
/help | Shows the available commands. | Use when you forget syntax or want a quick reminder of what the bot can do from chat. |
/commands | Lists all commands. | Use when you want the full command catalog instead of a short help view. |
/status | Checks token usage, model, context, and cache details. | Use regularly so you know your current session state before problems pile up. |
/reset | Clears the current conversation and starts fresh. You can optionally include a model hint. | Use when the session feels stuck, messy, or drifted too far off course. |
/new | Starts a fresh session, with an optional model hint if needed. | Use when you are changing topics and want a clean start instead of dragging old context with you. |
/model | Shows or changes the active model, including picker, number, or model-name selection. | Use to verify the model before deployments or switch to a model better suited for the task. |
/stop | Aborts the current response. | Use when the bot is going in the wrong direction, taking too long, or you want to interrupt a bad answer fast. |
Session & Context
| Command | What It Does | When To Use |
|---|---|---|
/compact | Forces context compaction. | Use when a session is getting heavy and you want to shrink context before it becomes a problem. |
/context | Shows what is currently in context, including files, tools, and system prompt sizes. | Use when you need to see what baggage is being carried into the next answer. |
/export-session | Exports the conversation to HTML. | Use when you want a clean archive, handoff, or record of an important session. |
/subagents list|kill|steer | Manages running sub-agents from one command family. | Use when you want to inspect, stop, or redirect specialist work already in progress. |
/kill <id|all> | Kills a specific sub-agent or all of them. | Use when a sub-agent is stuck, off-task, or no longer needed. |
/steer <id> <message> | Redirects a sub-agent with new instructions. | Use when the work is salvageable but needs a sharper brief or course correction. |
Behavior Directives
| Command | What It Does | When To Use |
|---|---|---|
/think <off|low|medium|high|xhigh> | Sets the thinking or reasoning level. | Use when you want to trade speed for more deliberate reasoning, or reduce reasoning to move faster. |
/fast on|off | Toggles fast mode, which skips some processing. | Use when you want quicker answers and can tolerate less ceremony or depth. |
/verbose on|full|off | Controls how detailed the output should be. | Use when debugging, tracing behavior, or when you want concise answers instead of fuller explanations. |
/reasoning on|off|stream | Shows reasoning in a separate message or turns it off. | Use for complex tasks where seeing the reasoning process helps, or disable it to keep chat cleaner. |
/elevated on|off|ask|full | Controls how exec permissions are handled. | Use when you want tighter safety controls, automatic prompting, or broader permission handling for command execution. |
Operational Commands
| Command | What It Does | When To Use |
|---|---|---|
/approve <id> allow-once|allow-always|deny | Approves or denies elevated commands when the system asks for permission. | Use when prompted to allow, permanently allow, or reject a sensitive command. |
/usage off|tokens|full|cost | Controls how usage information is shown after each response. | Use when you want lighter output or want to watch tokens and cost more closely. |
/tts off|always|inbound|tagged | Controls text-to-speech behavior. | Use when you want spoken replies always, only on inbound content, only when tagged, or not at all. |
/restart | Restarts the gateway. | Use when the gateway needs a clean restart after config changes or a bad state. |
/activation mention|always | Changes group activation mode. | Use in group chats to decide whether the bot should respond only when mentioned or always listen. |
Advanced Commands
| Command | What It Does | When To Use |
|---|---|---|
/config show|get|set|unset | Views or edits config from chat. Requires commands.config: true. | Use when you need to inspect or change configuration without leaving Telegram. |
/debug show|set|unset|reset | Manages runtime overrides. Requires commands.debug: true. | Use when diagnosing behavior and you need temporary overrides without editing files directly. |
/bash <command> | Runs a shell command from chat. Requires commands.bash: true. | Use for quick operational commands when chat control is faster than opening a shell. |
/skill <name> [input] | Runs a skill by name. | Use when a packaged workflow already exists for the job and you want to trigger it directly from chat. |
/queue <mode> | Manages message queue settings. | Use when you need to control how inbound work is buffered or processed. |
/whoami | Shows your sender ID. | Use when troubleshooting permissions, allowlists, routing, or identity-based settings. |
/allowlist | Manages access lists. | Use when controlling who can interact with the bot or use restricted commands. |
Telegram Tips & Gotchas
Every message is a separate prompt
Do not drip-feed half thoughts. Each new message can trigger the bot on its own.
Bots can't see bots
Do not expect one Telegram bot to read or react to another bot's messages in the chat.
Compose full thoughts before sending
Write the full request first. Clear prompts give better answers and reduce wasted back-and-forth.