ElevenLabs wired its whole voice agent platform into Claude through a hosted MCP connector. Install it from Claude’s connector directory, sign in with OAuth — no local server, no API key pasting — and start typing: spin up an agent, pick a voice, rewrite its system prompt. Done, without ever opening the ElevenLabs dashboard.
What the connector can actually do
This is a full read-write API surface over ElevenAgents, exposed as MCP tools. Claude can create agents, swap voices and languages, edit prompts and opening messages, pull conversation transcripts, check knowledge base size, compare configurations, and estimate LLM cost before applying a change. It can also delete a production agent — power and footgun in one sentence. Typical use case: “my support agent sounds robotic, make it warmer and tell me what the model switch costs” — one message, done.
Why this matters
The April 2025 version was a local server for speech generation. The 2026 hosted connector is different in kind: MCP graduating from “let Claude read data” to moving an entire third-party product’s control surface into a conversation. ElevenLabs buys distribution, Anthropic buys ecosystem lock-in, and the skill floor for building a voice agent drops to typing.
You Might Also Like
- Anthropic Agent sdk Metered Credits Openclaw Reinstated June 15 Splits Your Claude Subscription in two
- Claude for Microsoft 365 ga Outlook Beta Anthropic Puts one Agent Inside Word Excel Powerpoint and Outlook
- Claude Cowork Cloud Anthropic Close Your Laptop the Agent Keeps Working
- Anthropic cut 80 of Claude Codes System Prompt With Zero Eval Loss the new Rules of Context Engineering for Claude 5 Claude Doctor Doctor
- Anthropic Multi Agent Turf war Research 3 Claude Agents 1 Codebase Self Replicating Malware

Leave a comment