Skip to main content

Command Reference

All slash commands are sent directly to the agent via Telegram. They bypass the LLM and are handled synchronously by the event handler — responses are instant.

Basics

CommandDescription
/pingHealth check — responds pong
/statusFull system snapshot: LLM provider, active model, skills count, scheduler state, memory stats
/introspectDetailed health and performance report (CPU, RAM, error rates, skill latency)
/helpPrint all available commands

Model Management

CommandDescription
/model statusActive provider, model name, and configuration
/model listInstalled Ollama models
/model availableDownloadable models from Ollama library
/model switch <name>Switch active model for this session
/model default <name>Set persistent default model (survives restart)
/model download <name>Download a model in background; notifies when done
/model delete <name>Remove a locally stored model

Memory

CommandDescription
/memory statsEntry counts by type: facts, working, episodic, semantic, policy, meta
/memory recent [n]Last n episodic memories (default 5, max 10)
/memory search <text>Full-text search across all memory entries
/snapshot create <label>Save current cognitive state as a named snapshot
/snapshot listList all saved snapshots with timestamps

Scheduler & Tasks

CommandDescription
/schedule listAll 41 background jobs with interval and paused state
/schedule trigger <job>Run a scheduled job immediately
/schedule pause <job>Pause a job (skips future ticks until resumed)
/schedule resume <job>Resume a paused job

Custom recurring tasks (stored in DB, survive restarts):

CommandDescription
/task listAll custom tasks with status and last-run time
/task create <name> cada <interval>: <instruction>Create a recurring task
/task edit <name>Show task details and current instruction
/task trigger <name>Run a custom task immediately
/task pause <name>Pause a custom task
/task resume <name>Resume a paused custom task
/task delete <name>Permanently delete a custom task

Interval format examples: 1h, 30m, 6h, 1d

Skills

CommandDescription
/skillsFull skill list grouped by category, with ON/OFF status
/skill enable <name>Enable a skill
/skill disable <name>Disable a skill (agent cannot call it until re-enabled)

API Keys

CommandDescription
/api listAll configured providers and key status (masked)
/api set <provider> <key>Store an API key for a provider
/api remove <provider>Delete a provider's API key
/api test <provider>Run a live connection test against the provider

Supported providers: anthropic, openai, google, xai, mistral, cohere, fireworks, together, perplexity, groq, ollama

Web Monitoring

CommandDescription
/monitor listActive URL monitors with check interval and last status
/monitor remove <url|id>Remove a web monitor by URL or monitor ID

To add a new monitor, use the create_monitor skill via chat.

Agent Identity

CommandDescription
/identity showView the current identity prompt
/identity set <text>Update identity prompt (takes effect immediately, no restart needed)
/identity resetReset to built-in default identity
/identity versionsList all saved identity versions with timestamps
/identity rollback <timestamp>Restore a previous identity version

OpenClaw Skills

CommandDescription
/openclaw search <query>Search the ClawHub skill registry
/openclaw install <slug>Download and install a skill from ClawHub
/openclaw listAll installed OpenClaw skills
/openclaw remove <slug>Uninstall a skill

Infrastructure (Broker)

These commands route through the agent-broker privileged sidecar — the agent never has direct Docker socket write access.

CommandDescription
/broker status <container>Container state, uptime, and resource usage
/broker restart <container>Graceful restart of a container
/broker logs <container>Last 50 lines of container stdout

Available containers: agent-core, agent-telegram, agent-nginx, agent-redis, agent-postgres


Notes

  • Commands prefixed with / are not processed by the LLM — they are parsed directly by the event handler and return instantly.
  • Any other text sent to the bot is routed through the full agent pipeline (LLM + skills).
  • Unknown commands return: Unknown command: /xyz — Use /help for available commands.
  • Commands are case-insensitive (/PING = /ping).