Speakeroo speaking coach

Speakeroo MCP server

An AI speaking coach for any MCP client. Paste a speech, a talk or an interview answer, or send a recording, and get the feedback Speakeroo gives after a practice session: weak and filler words, rhetorical devices, grammar, vocabulary and coherence scores with fixes, a restructured version, and, from audio, pace, pauses, pitch and vocal expression.

• Endpoint: https://mcp.speakeroo.app/mcp (Streamable HTTP)
• Free, anonymous, no API key
• Registry name: app.speakeroo/speaking-coach (server.json)

This repository holds the public listing files for the hosted Speakeroo MCP server: install configs, server.json, llms-install.md and the logo. The server runs at https://mcp.speakeroo.app/mcp; there is nothing to build or run locally.

Install

Claude Desktop

Settings → Connectors → Add custom connector, then paste https://mcp.speakeroo.app/mcp.

Or add it to claude_desktop_config.json through mcp-remote:

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

Claude Code

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

Cursor

~/.cursor/mcp.json (or .cursor/mcp.json in a project):

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

LM Studio

Open lmstudio://add_mcp?config=eyJ1cmwiOiJodHRwczovL21jcC5zcGVha2Vyb28uYXBwL21jcCJ9&name=speakeroo to add it in one click, or add it to mcp.json:

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

VS Code

.vscode/mcp.json:

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

Codex

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

Or in ~/.codex/config.toml:

[mcp_servers.speakeroo]
url = "https://mcp.speakeroo.app/mcp"

Other MCP clients

Point any client that speaks Streamable HTTP at https://mcp.speakeroo.app/mcp. No API key or headers are needed. Clients configured with an mcpServers file take:

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

A client that only starts local (stdio) servers can bridge to it with npx -y mcp-remote https://mcp.speakeroo.app/mcp.

Tools

Tool	Input	Returns
analyze_transcript	text, 30–1,500 words; optional context, topic	Weak and filler words with firmer alternatives, rhetorical devices, delivery and rhetoric feedback, grammar/vocabulary/coherence scores (1–10), word-level issues with the excerpt and a fix
refine_transcript	text, 30–1,500 words; optional context, topic	The speech rewritten into a structure (PREP, STAR, problem–solution, rule of three…), a 1–10 score for how well the original followed it, what changed, and a delivery tip
analyze_speech	audioUrl (public https) or audioBase64; ≤2 min, ≤10 MB; mp3, m4a/mp4, wav, webm, ogg, flac	Everything analyze_transcript returns, plus what only audio shows: words per minute, pace, pauses, filler rate, pitch, vocal expression, and pacing/pronunciation/engagement scores
get_speech_topics	category: impromptu, persuasive, informative, presentation, ielts_part_1, ielts_part_2	Random topics, linking to the full list
how_to_pronounce	word	IPA, respelling, syllables and stress, sound tips, example sentences
get_speaking_rubric	–	A four-level speaking rubric: fluency, pronunciation, vocabulary range, task completion
list_filler_words	–	Common filler words grouped by what they do, and how to stop using them

context is one of general, impromptu, presentation, speech, interview, ielts, conversation. Every result has structured output (outputSchema) and readable text, and links to the matching page on speakeroo.app.

Limits

Per network, resetting at 00:00 UTC:

• 5 text analyses a day, shared by analyze_transcript and refine_transcript, at most 3 a minute
• 1 spoken analysis a day; a try that fails on our side does not count, up to 3 tries a day
• 60 reference lookups a minute

Privacy

Submitted text and audio are used to produce the response and are not stored. Audio uploaded as base64 is held only for the call and deleted when it ends. Your IP address is used only to apply the limits: we keep a salted hash of it that changes every day, and delete it after two days. We count tool usage anonymously (which tool, whether it worked, which client connected), never with your text, audio or IP. See the privacy policy.

Try it

Here's my answer to "Tell me about a time you handled conflict". Analyze it with Speakeroo as an interview answer and tell me the three things to fix first.

Restructure this toast with Speakeroo so it lands in under a minute.

Analyze this recording with Speakeroo: https://example.com/my-talk.m4a

Support

Questions or problems: [email protected] or speakeroo.app/mcp.