AutoGLM ASR MCP Server

MCP server for high-quality speech-to-text transcription using Zhipu AutoGLM ASR.

CN: 一个面向 Agent 的语音转文字 MCP 服务，支持长音频分块、上下文传递和时间戳分段。

For AI-oriented setup details, see AI_SETUP_GUIDE.md.

For AI Agents (TL;DR)

• Type: MCP Server

• Domain: ASR / speech-to-text / transcription

• Input: local audio file path

• Output: full transcript text + timestamp segments

• Best for: meeting notes, call analysis, subtitle draft, voice memo transcription

• Supported audio formats: mp3, wav, m4a, flac, ogg, webm

• Core tools: transcribe_audio, get_audio_info

Related MCP server: Voice to Text MCP Server

What It Does

• Transcribes short and long audio files with automatic chunking.

• Uses context-aware modes to balance speed and quality.

• Returns readable full text and segment-level timestamps.

• Runs over stdio as an MCP server for coding assistants.

Tool Index

Features

• Fast long-audio transcription with sliding-window concurrency.

• Better accuracy through chunk-to-chunk context passing.

• Automatic splitting for long inputs (API limit friendly).

• Zero-install runtime with npx.

• Works with common MCP clients.

Installation

Prerequisites

ffmpeg must be installed:

# macOS
brew install ffmpeg

# Ubuntu/Debian
apt install ffmpeg

# Windows
choco install ffmpeg

Get your API key from Zhipu AI Open Platform.

NPX (Recommended)

npx autoglm-asr-mcp

Quick Start

Add this MCP server to your client config and set AUTOGLM_ASR_API_KEY.

{
  "mcpServers": {
    "autoglm-asr": {
      "command": "npx",
      "args": ["-y", "autoglm-asr-mcp"],
      "env": {
        "AUTOGLM_ASR_API_KEY": "your-api-key"
      }
    }
  }
}

Compatibility

• Claude Desktop / Claude Code

• Cursor

• Windsurf

• VS Code MCP

• Other MCP-compatible clients

VS Code quick install:

Tools

transcribe_audio

Transcribe an audio file into text with timing segments.

Arguments:

Returns:

• Full transcription text

• Timestamped segment list

• Basic run stats (chunks, mode, elapsed time)

Common errors:

• File not found or unreadable path

• Unsupported format or broken audio stream

• Missing/invalid API key

get_audio_info

Inspect an audio file before transcription.

Arguments:

Returns:

• Duration

• Format

• Sample rate

• Channels

• Estimated chunks

Context Modes

Environment Variables

Use Cases

• Meeting recording to editable transcript

• Customer support call transcription

• Podcast/video subtitle draft generation

• Voice memo indexing and search

Limitations

• Requires local file path input (not remote URL input).

• Audio quality strongly affects transcription quality.

• Very noisy or multi-speaker overlap can reduce accuracy.

Troubleshooting

• ffmpeg not found: install ffmpeg and retry.

• File not found: pass an absolute existing path.

• API errors: verify AUTOGLM_ASR_API_KEY and account quota.

Keywords

mcp, model-context-protocol, asr, speech-to-text, transcription, autoglm, zhipu, chinese-asr, audio-transcription, meeting-transcript, subtitle-generation, voice-to-text, agent-tools, llm-tools, coding-agent

License

MIT