Tools & Plugins

Tools and Plugins

Everything the agent does beyond generating text happens through tools. Tools are how the agent reads files, runs commands, browses the web, sends messages, and interacts with devices.

​ agent 除生成文本外的所有操作都通过工具完成。工具是 agent 读取文件、运行命令、浏览网页、发送消息以及与设备交互的方式。

Tools, skills, and plugins

OpenClaw has three layers that work together:

​ OpenClaw 有三个协同工作的层级：

1. Tools are what the agent calls 工具是 agent 所调用的对象

A tool is a typed function the agent can invoke (e.g. exec, browser, web_search, message). OpenClaw ships a set of built-in tools and plugins can register additional ones. The agent sees tools as structured function definitions sent to the model API. 工具是智能体可调用的带类型函数（例如 exec、browser、web_search、message）。OpenClaw 自带一组 内置工具，插件可注册额外的工具。 智能体将工具视为发送至模型 API 的结构化函数定义。

2. Skills teach the agent when and how - Skills 教会智能体何时以及如何行动

A skill is a markdown file (SKILL.md) injected into the system prompt. Skills give the agent context, constraints, and step-by-step guidance for using tools effectively. Skills live in your workspace, in shared folders, or ship inside plugins.

​ 技能是一个注入到系统提示词中的 Markdown 文件（SKILL.md）。技能为智能体提供上下文、约束条件，以及有效使用工具的分步指导。技能可以存放在你的工作区、共享文件夹中，也可以内置在插件里。

Skills reference | Creating skills

3. Plugins package everything together 插件将所有内容整合在一起

A plugin is a package that can register any combination of capabilities: channels, model providers, tools, skills, speech, image generation, and more. Some plugins are core (shipped with OpenClaw), others are external (published on npm by the community).

​ 插件是一个可以注册任意功能组合的包：channels、模型提供商、工具、skills、语音、图像生成等等。一些插件是核心插件（随 OpenClaw 一同发布），另一些是外部插件（由社区发布在 npm 上）。

Install and configure plugins | Build your own

Built-in tools 内置工具

These tools ship with OpenClaw and are available without installing any plugins:

​ 这些工具随 OpenClaw 一同提供，无需安装任何插件即可使用：

For image work, use image for analysis and image_generate for generation or editing. If you target openai/*, google/*, fal/*, or another non-default image provider, configure that provider’s auth/API key first.

​ 对于图像相关操作，使用 image 进行分析，使用 image_generate 进行生成或编辑。如果你的目标是 openai/*、google/*、fal/* 或其他非默认图像提供商，请先配置该提供商的身份auth/API key。

Plugin-provided tools 插件提供的工具

Plugins can register additional tools. Some examples:

​ 插件可以注册额外的工具。以下是一些示例：

• Lobster — typed workflow runtime with resumable approvals 支持可恢复审批的强类型工作流运行时
• LLM Task — JSON-only LLM step for structured output 仅输出 JSON 格式的大语言模型步骤，用于生成结构化结果
• Diffs — diff viewer and renderer 差异查看与渲染工具
• OpenProse — markdown-first workflow orchestration 以Markdown为核心的工作流编排

Tool configuration

Allow and deny lists 允许列表与拒绝列表

Control which tools the agent can call via tools.allow / tools.deny in config. Deny always wins over allow.

​ 在配置中通过 tools.allow / tools.deny 控制智能体可调用的工具。拒绝规则的优先级始终高于允许规则。

1
2
3
4
5
6

{
  tools: {
    allow: ["group:fs", "browser", "web_search"],
    deny: ["exec"],
  },
}

Tool profiles 工具配置

tools.profile sets a base allowlist before allow/deny is applied. Per-agent override: agents.list[].tools.profile.

​ tools.profile 在应用 allow/deny 之前设置基础允许列表。单代理覆盖：agents.list[].tools.profile。

Tool groups 工具组

Use group:* shorthands in allow/deny lists:

​ 在允许/拒绝列表中使用 group:* 简写：

Provider-specific restrictions 提供商特定限制

Use tools.byProvider to restrict tools for specific providers without changing global defaults:

​ 使用 tools.byProvider 来限制特定提供商的工具，而不更改全局默认值：

1
2
3
4
5
6
7
8

{
  tools: {
    profile: "coding",
    byProvider: {
      "google-antigravity": { profile: "minimal" },
    },
  },
}