DevOps / SRE / Platform · 17.08.2026, 20:10 UTC
Claude can now delete your production voice agent from a chat window
| Schweregrad | info |
|---|---|
| Kategorie | DevOps / SRE / Platform |
| Quelle | The New Stack ↗ |
| Veröffentlicht | 17.08.2026 UTC |
Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.
Developers can now ask Claude to inspect a production voice agent, revise its system prompt, change the voice entirely, or estimate the cost of using a different language model without ever opening the ElevenLabs dashboard.
Monday, ElevenLabs introduced the hosted MCP connector, allowing Claude read and write access to chat agents built with ElevenAgents. In addition to creating agents and comparing configurations, the connector can calculate expected LLM usage and cost before a change is applied. For example, a developer could ask Claude: “What would my checkout agent cost per conversation with Gemini 2.5 Flash instead of GPT-4o?” This type of question is even more important now that cheaper models alone don’t guarantee lower bills once agentic orchestration is involved.
“What would my checkout agent cost per conversation with Gemini 2.5 Flash instead of GPT-4o?”
OAuth replaces local API keys
ElevenLabs released its original open-source server in April 2025 for Claude Desktop, Cursor, and other MCP clients. Developers ran it locally, added an ElevenLabs API key and used it to generate speech, clone voices or transcribe audio. The new server focuses on managing agents that already exist in an ElevenLabs workspace.
Developers install the connector from Claude’s directory and sign in via OAuth, eliminating the need to run a server or paste an API key into Claude, while limiting access to the ElevenLabs workspace and to permissions approved during login.
According to ElevenLabs, Claude can update an agent’s system prompt, language, voice, and opening message; retrieve …