> This is a page from the ElevenLabs documentation. For a complete page index, fetch https://elevenlabs.io/docs/llms.txt. For the full documentation in a single file, fetch https://elevenlabs.io/docs/llms-full.txt.

# 智能体工具

## 概览

智能体工具包括可复用技能和本地集成，可帮助你在开发者工具和助手中自动化 ElevenLabs workflow。

## 智能体技能

智能体技能是用于常见 ElevenLabs workflow 的可复用构建模块。它们遵循智能体技能规范，可与兼容的编程助手配合使用。使用以下命令安装技能：

```bash
npx skills add elevenlabs/skills
```

官方合集托管在 [elevenlabs/skills](https://github.com/elevenlabs/skills) 仓库中。
安装技能后，打开各个技能即可查看可粘贴到助手中的提示词。

#### 文本转语音

使用 ElevenLabs 音色将文本转换为语音。

```text
Use the text to speech skill to generate audio for the script below.
Model: eleven_v4
Voice: "Juniper" (or default).
Output: Save the MP3 file locally and return the file path.
Script: "Welcome to ElevenLabs. Today we will walk through the new agent tooling."
```

#### 语音转文本

将音频转录为带时间戳的文本。

```text
Use the speech to text skill to transcribe the audio file path/to/file.mp3.
Return a transcript that contains speaker IDs and timestamps at the start of each paragraph.
```

#### 实时语音转文本

以低延迟流式传输实时转录内容。

```text
Use the speech to text skill to start a real-time transcription session for microphone input.
Model: scribe_v2_realtime.
Stream partial transcripts and return committed transcripts with word-level timestamps.
```

#### 智能体

构建对话式语音智能体。

```text
Use the Agents skill to create a voice agent named "Support Concierge".
Persona: friendly, concise, asks clarifying questions when needed.
Goals: answer pricing questions and route enterprise leads to sales.
```

#### Speech Engine

使用 [Speech Engine 技能](https://github.com/elevenlabs/skills/tree/main/speech-engine)为自定义聊天智能体添加语音功能。

```text
Use the Speech Engine skill to add voice to my chat agent.
Server: Node.js with an existing HTTP server.
LLM: Use my existing chat completion logic.
Output: Create the Speech Engine server route, client connection code, and setup steps.
```

#### 音效

根据文本提示词生成音效。

```text
Use the Sound effects skill to generate a 3-second effect: "Wooden door creaks open, then a soft slam."
Output WAV at 48 kHz and return the file path.
```

#### 音乐

根据提示词生成音乐曲目。

```text
Use the Music skill to generate a 30-second instrumental loop.
Style: lo-fi hip hop, warm, chill.
BPM: 80. No vocals.
Return the audio file path.
```

#### 变声器

在保留原始表达方式的同时，将录音转换为不同的目标音色。

```text
Use the voice changer skill to transform path/to/file.mp3 into the voice "George".
Model: eleven_multilingual_sts_v2.
Return the transformed audio file path.
```

#### 人声分离

移除音频文件中的背景噪声。

```text
Use the voice isolator skill to remove background noise from path/to/file.mp3.
Return the cleaned audio file path.
```

#### 设置 API 密钥

获取并配置 ElevenLabs API 密钥。

```text
How do I get my ElevenLabs API key?
```

## ElevenLabs MCP 服务器

托管 MCP 服务器是一种远程 Model Context Protocol 服务器，Claude 等助手可通过简单提示词管理你的智能体。它使用 ElevenLabs 账户进行 OAuth 登录，无需本地安装。可从 Claude Desktop 目录连接。

#### [托管 MCP 服务器](/docs/zh/eleven-agents/operate/hosted-mcp)

从 Claude Desktop 目录连接，无需安装。