Best for
- Developers and teams wanting to integrate text-to-speech or audio processing into their applications
- Users of MCP clients like Claude Desktop who need high-quality voice generation
E
Elevenlabs MCP
Official
Archived
Official ElevenLabs MCP server for Text to Speech and audio processing
- Category
- Other
- Stars
- ★ 1.5k Very popular
- Transport
- stdio (local process)
- Runtime
- Python · Prebuilt binary
- Credentials
- API key / credential required
- License
- MIT
- Last commit
66FMRS · C
This server is an official MCP integration by ElevenLabs, powerful but requires API key and credit usage. Configuration is straightforward and compatible with Claude Desktop. Suitable for users needing high-quality speech synthesis.
Strongest · Documentation 15/20 Weakest · Reliability 10/20
Why each score
- Reliability 10/20
- The README provides a concrete happy path (uvx elevenlabs-mcp; python -m elevenlabs_mcp --api-key ... --print) and names specific tools (speech_to_text, isolate_audio, speech_to_speech, video_to_music, upload_music_for_inpainting) plus a resource URI scheme (elevenlabs://filename), so a clean start and tool-list handshake are plausible. However, the supplied material contains only the README and manifest — no CI workflow files, test files, or executable source to verify — and the manifest is internally inconsistent (server version 0.9.1 vs package version 0.8.1), while the Tests badge points at a differently named repository (elevenlabs/elevenlabs-mcp-server). Under the static rules reliability is capped at 12 without verifiable execution evidence; the version contradiction and missing evidence reduce it to 10.
- Security and permissions 14/20
- Credentials are handled via a required secret environment variable (ELEVENLABS_API_KEY), and the install example uses a placeholder rather than a real token, so no red-line issue (malware, credential theft, covert exfiltration, or real secrets in docs) is present. The base path ELEVENLABS_MCP_BASE_PATH doubles as an input-file sandbox: paths outside it are rejected, which is a clearly disclosed data boundary; output modes (files/resources/both) and disk side effects are documented, and the README explicitly warns that tools consume ElevenLabs credits. Deductions: no per-operation confirmation for credit-consuming calls, no documentation of API-key scope (least privilege), a broad default write location of ~/Desktop, and data residency marked as an enterprise-only feature without retention/sharing detail.
- Maintenance 13/20
- The repository is not archived, has an MIT license, has clear ownership (official ElevenLabs), publishes versioned packages on PyPI, shows only 8 open issues, declares status active in the manifest, and the README includes a contribution guide and community channels. Deductions: the supplied material contains no commit/release history, changelog, or dependency-update evidence, there is no SECURITY.md or security-response channel, and the manifest/package version mismatch (0.9.1 vs 0.8.1) is a governance gap. Stars were treated as discovery signals only and added no points.
- Documentation 15/20
- The README is reasonably layered: a Claude Desktop quickstart with a ready JSON config block, an alternative pip/python flow for other clients, example prompts, three output modes with config examples and resource-mode benefits, data residency, contributing steps, and a troubleshooting section with per-OS log paths, timeout notes, and the uvx ENOENT fix. Limitations such as credit costs, long-running operations, and Windows Developer Mode are disclosed. Deductions: the supplied material lacks a per-tool parameter reference, rate limits are not covered, the Tests badge links to a repository with a different name, and the troubleshooting section is thin with only two entries.
- Setup experience 14/20
- Setup is a short, concrete path: install uv, then paste a JSON block with a placeholder key into Claude Desktop, or for Cursor/Windsurf run pip install plus python -m elevenlabs_mcp --api-key ... --print to obtain configuration. Prerequisites (uv, API key, free tier) and a Windows-specific note are provided. The static cap is 15 because no verifiable execution evidence (CI/test files) is present in the supplied material; the documented uvx ENOENT failure mode, the manual config copying required for other clients, the Windows Developer Mode requirement, and the lack of alternative deployment options (e.g., Docker) keep it below the cap.
Static review · not runListed 2026-08-07
Read the FMRS scoring method →Fit and risk
What it can accessReads local filesWrites / deletes local filesUses the network
Not for
- Users without need for audio processing
- Users preferring other TTS providers
- Scenarios requiring real-time streaming audio (may not be supported)
Required permissions
- Requires ElevenLabs API key (mandatory)
- Requires file system read/write access (defaults to Desktop, configurable with ELEVENLABS_MCP_BASE_PATH)
- Optional ELEVENLABS_MCP_OUTPUT_MODE to control output (files or resources)
Risks and side effects
- API usage consumes credits; free tier has limits, monitor usage
- API key leakage may lead to unauthorized use, keep it secure
- Audio files may contain sensitive information, ensure data privacy
- Some operations (e.g., audio isolation) may take long and cause timeouts
Setup
Before you start
Runtime：Python · Prebuilt binary
ELEVENLABS_API_KEY
Required ElevenLabs API key; get it at elevenlabs.io/app/settings/api-keys (free tier: 10k credits/month).
Other optional settings (3)
ELEVENLABS_MCP_BASE_PATH
Optional; base directory for file operations (defaults to $HOME/Desktop) and the security boundary for reading local files.
ELEVENLABS_API_RESIDENCY
Optional; data residency region, defaults to 'us' (enterprise-only feature).
ELEVENLABS_MCP_OUTPUT_MODE
Optional; output mode, defaults to 'files'; supports 'resources' and 'both'.
- Get your API key from ElevenLabs (free tier with 10k credits per month).
- Install uv: curl -LsSf https://astral.sh/uv/install.sh | sh
- Configure Claude Desktop by adding the MCP server to claude_desktop_config.json with command 'uvx', args ["elevenlabs-mcp"], and environment variable ELEVENLABS_API_KEY.
Check that it works
Add the ElevenLabs entry to Claude Desktop's mcpServers config with your ELEVENLABS_API_KEY, restart, then ask Claude to generate a thunderstorm soundscape — a returned audio file or resource confirms the server is connected.
Troubleshooting
- Ensure ELEVENLABS_API_KEY is set correctly
- If encountering 'spawn uvx ENOENT', find the absolute path of uvx and update config
- Check logs: Windows: %APPDATA%\Claude\logs\mcp-server-elevenlabs.log; macOS: ~/Library/Logs/Claude/mcp-server-elevenlabs.log
- Ensure file paths are within ELEVENLABS_MCP_BASE_PATH directory to avoid rejection
Things to try
Once connected, you can ask your AI assistant things like:
- Create an AI agent that speaks like a film noir detective and can answer questions about classic movies
- Generate three voice variations for a wise, ancient dragon character
- Convert this recording of my voice to sound like a medieval knight
- Create a soundscape of a thunderstorm in a dense jungle with animals reacting to the weather
Use cases
Generate speech from text (Text to Speech)
Clone voices for unique characters
Create interactive voice agents
Transcribe audio to text (Speech to Text)
Convert audio (e.g., speech-to-speech, video-to-music)
Isolate audio (separate vocals and music)
Supported clients
Claude Desktop
Listed from the project's documentation, not tested by this site.
Overview
The official ElevenLabs Model Context Protocol (MCP) server enables interaction with powerful Text to Speech and audio processing APIs. It allows MCP clients like Claude Desktop, Cursor, Windsurf, and OpenAI Agents to generate speech, clone voices, transcribe audio, and more.
Similar servers
DuckDuckGo Search MCP Server
74 · B
MCP server providing web search and content fetching via DuckDuckGo
Basic Memory
72 · B
Your AI never forgets again.
Telegram MCP Server
70 · B
Telegram MCP server powered by Telethon to let MCP clients read chats, manage groups, and send/modify messages, media, contacts, and settings.
Hostinger API MCP Server
70 · B
Official MCP server for Hostinger API.
Source revision c6dc1688ead5 Data synced 2026-10-12 Read the FMRS scoring method