Speakeroo speaking coach

AI演讲教练：提供对文本或音频演讲的反馈，以及主题、发音和评分标准。

访问 speakeroo.app

托管 MCP 服务器

npx add-mcp 'https://mcp.speakeroo.app/mcp'

可安装到 Claude Code、Codex、Cursor 等客户端

文档

Speakeroo MCP server

A free speaking coach for your AI assistant. Paste a transcript or share a short recording in Claude, Cursor or VS Code and get feedback on filler words, structure, grammar, pace and delivery. No account and no API key: add one URL and start asking.

https://mcp.speakeroo.app/mcp

Streamable HTTP. No API key, no account.

Practise out loud on Speakeroo

Connect it to your AI client

Claude Desktop

claude_desktop_config.json

{
  "mcpServers": {
    "speakeroo": {
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.speakeroo.app/mcp"
      ],
      "command": "npx"
    }
  }
}

Claude Desktop's config file starts local servers, so mcp-remote bridges it to the remote endpoint. You can also skip the file: open Settings → Connectors → Add custom connector and paste https://mcp.speakeroo.app/mcp.

Claude Code

Terminal

claude mcp add --transport http speakeroo https://mcp.speakeroo.app/mcp

Cursor

~/.cursor/mcp.json

Add to Cursor

{
  "mcpServers": {
    "speakeroo": {
      "url": "https://mcp.speakeroo.app/mcp"
    }
  }
}

VS Code

.vscode/mcp.json

Install in VS Code Install in VS Code Insiders

{
  "servers": {
    "speakeroo": {
      "type": "http",
      "url": "https://mcp.speakeroo.app/mcp"
    }
  }
}

LM Studio

mcp.json

Add to LM Studio

{
  "mcpServers": {
    "speakeroo": {
      "url": "https://mcp.speakeroo.app/mcp"
    }
  }
}

Codex

Terminal

codex mcp add speakeroo --url https://mcp.speakeroo.app/mcp

Or add it to ~/.codex/config.toml as [mcp_servers.speakeroo] with url = "https://mcp.speakeroo.app/mcp".

Other MCP clients

Any MCP client

{
  "mcpServers": {
    "speakeroo": {
      "url": "https://mcp.speakeroo.app/mcp"
    }
  }
}

Most clients take the endpoint https://mcp.speakeroo.app/mcp with the Streamable HTTP transport, or this mcpServers entry in their config file. No API key or headers are needed. A client that only starts local servers can bridge to it with npx -y mcp-remote https://mcp.speakeroo.app/mcp.

Analysis tools

analyze_transcript

Send 30 to 1,500 words you plan to say. You get weak and filler words with firmer alternatives, the rhetorical devices you used, delivery and rhetoric feedback, 1–10 scores for grammar, vocabulary and coherence, and word-level issues, each with the excerpt and a fix. Optionally set the context (general, impromptu, presentation, speech, interview, ielts or conversation) and a topic.

refine_transcript

Send a transcript and get it back rewritten into a clear structure such as PREP, STAR, problem–solution or the rule of three, with a 1–10 structure score, a summary of what changed and one delivery tip.

analyze_speech

Send a recording as a public https URL or base64 audio: up to 2 minutes and 10 MB, in mp3, m4a/mp4, wav, webm, ogg or flac. You get everything analyze_transcript returns, plus what only audio shows: words per minute, pace, pauses, filler rate, pitch, vocal expression, and pacing, pronunciation and engagement scores.

Free reference tools

get_speech_topics

Random topics to practise with: impromptu, persuasive, informative, presentation, IELTS Speaking Part 1 questions or Part 2 cue cards. Impromptu speech topics

how_to_pronounce

IPA, a plain respelling, syllables with the stress marked, sound tips, the usual mistake and two example sentences for an English word. How to pronounce

get_speaking_rubric

A four-level rubric for fluency, pronunciation, vocabulary range and task completion, with a descriptor for every level. Speaking rubric

list_filler_words

Common filler words in spoken English grouped by what they do, and the steps to stop using them. Filler words

Things to ask

  • Analyse this interview answer with Speakeroo and tell me which filler words to cut first: …
  • Use Speakeroo to refine my wedding toast with the rule of three, then give me the delivery tip.
  • Analyse the recording at https://example.com/pitch.m4a and tell me my words per minute and where I paused too long.
  • Give me three persuasive speech topics from Speakeroo, then score my answer to the first one when I paste it.
  • How do I pronounce "specificity"? Use the Speakeroo tool.
  • Fetch the Speakeroo speaking rubric and use it to grade this IELTS Part 2 answer: …

Limits

Text analyses

5 a day, shared by analyze_transcript and refine_transcript, and at most 3 a minute.

Spoken analysis

1 analyze_speech result a day. A try that fails on our side gives it back, up to 3 tries a day.

Reference tools

60 calls a minute for get_speech_topics, how_to_pronounce, get_speaking_rubric and list_filler_words.

Per network, daily

Every limit is free, counted per network, and resets at 00:00 UTC.

Privacy

  • Text and audio you send are not stored after the response.
  • Audio is deleted as soon as it has been analysed.
  • Your IP address is only used as a salted hash that rotates daily, to count limits, and that hash is deleted after two days.
  • Usage is counted anonymously (which tool, whether it worked), never with your text, audio or IP.
  • No account, no sign-in and no API key.

Frequently asked questions

Is the Speakeroo MCP server free?

Yes. Every tool is free with no account: 5 text analyses and 1 spoken analysis a day per network, plus the reference tools. Limits reset at 00:00 UTC.

Do I need an API key?

No. The server at https://mcp.speakeroo.app/mcp is anonymous over Streamable HTTP, so you add the URL to your client and call the tools.

What's the difference between analyze_transcript and analyze_speech?

analyze_transcript reads text, so it judges your words, structure, grammar and vocabulary. analyze_speech listens to a recording, so it adds what only audio shows: words per minute, pauses, filler rate, pitch, vocal expression and pronunciation.

Is my audio stored?

No. Audio is deleted once it has been analysed, and neither text nor audio is stored after the response. Your IP address is only kept as a salted, daily-rotating hash for limits, deleted after two days.

What are the limits?

5 text analyses a day shared by analyze_transcript and refine_transcript (at most 3 a minute), 1 spoken analysis a day, and 60 reference calls a minute. Limits are per network and reset at 00:00 UTC. A spoken analysis that fails on our side does not count, up to 3 tries a day.

Which AI clients does it work with?

Any MCP client that speaks Streamable HTTP, including Claude Code, Cursor and VS Code. Claude Desktop connects through mcp-remote or a custom connector.

Keep going

Filler words Speaking rubric Impromptu speech topics How to pronounce Words per minute speaking AI speaking coach

Hear how it sounds out loud

Every practice mode is free and starts in your browser. A free account adds session history and progress tracking so you can see the same feedback improve over time.

Start a quick session Sign up for free

Get the app