Linux command
zeroclaw 命令
文本
复制后可按需替换文件名、目录或参数。
常用示例
Initialize
zeroclaw onboard --interactive
Start an interactive
zeroclaw agent
Send a single message
zeroclaw agent -m "[message]"
Start the full autonomous daemon
zeroclaw daemon
Check system health
zeroclaw doctor
List all supported
zeroclaw providers
说明
zeroclaw is a lightweight, security-first autonomous AI agent runtime built entirely in Rust. It serves as infrastructure for agentic workflows, abstracting models, tools, memory, and execution into a single binary that can be deployed across ARM, x86, and RISC-V architectures. The runtime compiles to an approximately 9MB binary with cold start under 10 milliseconds and less than 5MB RAM usage at idle. It uses a trait-driven architecture where providers, channels, memory backends, and tools are swappable through configuration without code changes. ZeroClaw includes a built-in hybrid search memory system combining vector embeddings with keyword search (SQLite-backed), requiring no external dependencies. It supports 28+ LLM providers including OpenRouter, Anthropic, OpenAI, Gemini, Ollama, and any OpenAI-compatible endpoint. Messaging integrations cover 70+ channels including Telegram, Discord, Slack, iMessage, Matrix, Signal, and WhatsApp. The agent supports multi-turn conversations with context preservation, tool execution (shell, file, git, browser), scheduled tasks via cron expressions, and hardware peripheral control for IoT devices.
参数
- onboard
- Initialize or reconfigure the workspace. Creates ~/.zeroclaw/config.toml and scaffold files.
- onboard --interactive
- Run the full 9-step configuration wizard.
- onboard --api-key _KEY_ --provider _PROVIDER_ --model _MODEL_
- Non-interactive setup with API key and provider in one command.
- onboard --channels-only
- Repair or reconfigure channels and allowlists only.
- agent -m _MESSAGE_
- Run interactive CLI agent or send a single message with -m.
- agent --provider _PROVIDER_
- Override the default provider for this session.
- gateway --port _PORT_
- Start the HTTP/WebSocket server for external integrations. Default port: 42617.
- daemon
- Start the full autonomous runtime with gateway, channels, heartbeat, and scheduler.
- status
- Display comprehensive system status including provider, memory, channels, and security.
- doctor models | traces
- Run system diagnostics. Optionally check model catalogs or inspect runtime traces.
- service install | start | stop | status | restart
- Manage zeroclaw as a background system service (systemd or OpenRC).
- channel list | doctor | start
- List, health-check, or start configured messaging channels.
- channel add _TYPE_ _JSON_CONFIG_
- Add a new messaging channel configuration.
- channel bind-telegram _USER_ID_
- Add a Telegram user to the allowlist.
- auth login _PROVIDER_ _PROFILE_
- Authenticate via OAuth or device code flow.
- auth status | logout _PROFILE_ID_
- Show or remove authentication profiles.
- memory stats | list | search _QUERY_ | delete _KEY_ | prune
- Manage the built-in memory system.
- cron add _NAME_ _SCHEDULE_ _MESSAGE_ --tz _TIMEZONE_
- Schedule a recurring task with a cron expression.
- cron list | remove _NAME_
- List or remove scheduled tasks.
- skills list | install _NAME_ | remove _NAME_
- Manage agent skills and extensions.
- migrate openclaw --dry-run
- Import memory from an OpenClaw runtime.
- completions _SHELL_
- Generate shell completions for bash, zsh, fish, powershell, or nushell.
- providers
- List all supported LLM providers and aliases.
- estop --resume
- Engage or resume from emergency stop.
FAQ
What is the zeroclaw command used for?
zeroclaw is a lightweight, security-first autonomous AI agent runtime built entirely in Rust. It serves as infrastructure for agentic workflows, abstracting models, tools, memory, and execution into a single binary that can be deployed across ARM, x86, and RISC-V architectures. The runtime compiles to an approximately 9MB binary with cold start under 10 milliseconds and less than 5MB RAM usage at idle. It uses a trait-driven architecture where providers, channels, memory backends, and tools are swappable through configuration without code changes. ZeroClaw includes a built-in hybrid search memory system combining vector embeddings with keyword search (SQLite-backed), requiring no external dependencies. It supports 28+ LLM providers including OpenRouter, Anthropic, OpenAI, Gemini, Ollama, and any OpenAI-compatible endpoint. Messaging integrations cover 70+ channels including Telegram, Discord, Slack, iMessage, Matrix, Signal, and WhatsApp. The agent supports multi-turn conversations with context preservation, tool execution (shell, file, git, browser), scheduled tasks via cron expressions, and hardware peripheral control for IoT devices.
How do I run a basic zeroclaw example?
Run `zeroclaw onboard --interactive` in a terminal, then adjust file names, paths, flags, or remote targets for your system.
What does onboard do in zeroclaw?
Initialize or reconfigure the workspace. Creates ~/.zeroclaw/config.toml and scaffold files.