ElevenLabs

Build and operate ElevenAgents voice and chat agents: prompts, voices, knowledge bases, branches, tests and triage…

Add to Agentman Studio

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

Resources

What the ElevenLabs connector does

The ElevenLabs MCP server connects your ElevenAgents workspace to Claude, ChatGPT and any MCP-compatible agent. This is agent operations, not text-to-speech: its 79 tools create and configure voice and chat agents, manage their knowledge bases and procedures, branch and deploy configuration changes, run tests, and read conversation transcripts. ElevenLabs hosts it, so nothing installs locally.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

ElevenLabs tools (79)

  • agents_add_triage_ticket_comment
  • agents_add_triage_ticket_turn_comment
  • agents_bulk_delete_knowledge_base
  • agents_bulk_move_knowledge_base
  • agents_calculate_llm_usage
  • agents_compile_procedures
  • agents_create
  • agents_create_branch
  • agents_create_deployment
  • agents_create_draft
  • agents_create_kb_folder
  • agents_create_kb_text
  • agents_create_kb_url
  • agents_create_manual_triage_ticket
  • agents_create_mcp_server
  • agents_create_procedure
  • agents_create_test
  • agents_create_tool
  • agents_create_triage_ticket
  • agents_delete
  • agents_delete_draft
  • agents_delete_kb_document
  • agents_delete_mcp_server
  • agents_delete_phone_number
  • agents_delete_procedure
  • agents_delete_procedure_draft
  • agents_delete_test
  • agents_delete_tool
  • agents_delete_triage_ticket
  • agents_duplicate
  • agents_get
  • agents_get_branch
  • agents_get_conversation
  • agents_get_kb_dependents
  • agents_get_kb_document
  • agents_get_knowledge_size
  • agents_get_link
  • agents_get_mcp_server
  • agents_get_phone_number
  • agents_get_procedure
  • agents_get_procedure_draft
  • agents_get_summaries
  • agents_get_test
  • agents_get_test_run
  • agents_get_tool
  • agents_get_tool_dependents
  • agents_get_tool_executions
  • agents_get_topics
  • agents_get_triage_ticket
  • agents_get_triage_ticket_assignable_users
  • agents_get_version
  • agents_get_widget
  • agents_list
  • agents_list_branches
  • agents_list_conversations
  • agents_list_knowledge_base
  • agents_list_mcp_server_tools
  • agents_list_mcp_servers
  • agents_list_phone_numbers
  • agents_list_procedures
  • agents_list_test_runs
  • agents_list_tests
  • agents_list_tools
  • agents_list_triage_tickets
  • agents_merge_branch
  • agents_merge_branch_preview
  • agents_query_knowledge_base_rag
  • agents_resolve_conversation
  • agents_run_tests
  • agents_search_conversation_messages
  • agents_search_knowledge_base
  • agents_update
  • agents_update_branch
  • agents_update_kb_document
  • agents_update_mcp_server
  • agents_update_phone_number
  • agents_update_procedure_draft
  • agents_update_tool
  • agents_update_triage_ticket

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • Knowledge base size is gated by subscription tier. ElevenLabs caps the total original file size a workspace can index for RAG at 1MB on Free, 2MB on Starter, 20MB on Creator, 100MB on Pro, 500MB on Scale and 1GB on Business and Enterprise, and notes that Free-tier indexes may be deleted after inactivity. Individual uploaded files are capped at 20MB, a document only fits in full context up to roughly 300,000 extracted characters, and documents under 500 bytes cannot be indexed for RAG at all.
  • Concurrency is a subscription limit, and exceeding it costs double. ElevenLabs' burst pricing documentation says burst calls are accepted up to three times your normal concurrency limit or 300, whichever is lower, are charged at twice the standard rate, are deprioritised for speech processing, and that calls beyond the burst limit are rejected with an error. Non-enterprise customers cannot exceed 300 concurrent calls.
  • LLM cost is metered and model-dependent. agents_calculate_llm_usage estimates it, but ElevenLabs' LLM pricing page states its listed prices are per million tokens, were accurate at the time of writing, and are subject to change by the providers.
  • The default endpoint is global only. A workspace in an EU, India or Singapore data residency environment needs a custom connector on that region's URL, and data residency is an Enterprise feature.
  • Outbound MCP is off by default and blocked on some plans. Before agents_create_mcp_server works, a workspace member must accept ElevenLabs' MCP Server Terms, and ElevenLabs states that MCP is unavailable to workspaces on Zero Retention Mode or requiring HIPAA compliance regardless of the can_use_mcp_servers setting.
  • Versioning is one-way. ElevenLabs states that once versioning is enabled on an agent it cannot be disabled, so the branch and deployment tools change an agent permanently the first time they are used on it.
  • We could not read the safety annotations. The server requires authentication, so the read and acts labels on this page come from ElevenLabs' documentation and API reference rather than from readOnlyHint or destructiveHint values we observed.
  • No published rate limits for the MCP endpoint. Our 401 responses on 2026-09-08 carried no x-ratelimit-* headers, and ElevenLabs' hosted MCP page states none. Assume the platform's normal API limits apply.

Frequently asked questions

Is the ElevenLabs MCP server for text-to-speech or for agents?

For agents. All 79 tool names in Anthropic's directory listing begin with the agents_ prefix and manage ElevenAgents resources — agent configuration, knowledge bases, procedures, branches, tests and triage tickets. ElevenLabs describes the server as exposing agent management tools, so an assistant can create, configure and manage the agents in your workspace.

Does the ElevenLabs MCP server use OAuth or an API key?

OAuth, and no API key is involved. ElevenLabs states that no API keys are copied into the client. Our anonymous request to the endpoint on 2026-09-08 returned HTTP 401 with a WWW-Authenticate Bearer challenge, and the server's own OAuth descriptor advertises seven scopes including convai_read and convai_write. You authorise once and pick a workspace.

Can the ElevenLabs connector delete my production voice agents?

Yes, if you approve the call. ElevenLabs states plainly that deleting an agent is destructive and advises reviewing tool calls in your MCP client before approving them, and restricting write access to workspace members who need it. Eleven of the 79 listed tools delete something, and one of those deletes knowledge base documents in bulk.

What URL do I use if my ElevenLabs workspace is in the EU?

Add a custom connector pointing at your region's URL instead of the default. ElevenLabs publishes separate MCP endpoints for EU, India and Singapore data residency environments. Those isolated environments are separate workspaces with separate accounts, so sign in with the account for that environment rather than your elevenlabs.io account.

How large a knowledge base can the ElevenLabs connector manage?

That depends on your subscription tier, not on the connector. ElevenLabs caps the total original file size indexed for RAG per workspace: 1MB on Free, 2MB on Starter, 20MB on Creator, 100MB on Pro, 500MB on Scale, and 1GB on Business and Enterprise. Individual uploaded files are capped at 20MB each.

Can I stop the assistant from running certain ElevenLabs tools?

Yes, in your MCP client rather than in ElevenLabs. ElevenLabs documents that clients such as Claude let you enable or disable individual tools and set whether each runs automatically or asks first. Workspace administrators can configure this for a whole organisation, and a tool an administrator disables cannot be re-enabled by an individual user.

Sources

More Developer Tools connectors

  • Base44

    Build and edit Base44 apps from an AI assistant, and query data your app's users submit.

  • CircleCI

    Diagnose failed builds, read job logs, inspect test results and rerun or roll back deployments from your AI assistant.

  • Claude Managed Agents

    Create, run, monitor and debug Anthropic's Claude Managed Agents from inside Claude itself.

Browse all MCP connectorsClaude and ChatGPT setup guideBrowse AI skills

Product names and logos belong to their owners. Agentman is not affiliated with a connector's maker unless the page says the connector is built and run by Agentman.

Use it in an agent

Put ElevenLabs to work.

Connect ElevenLabs once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Open in Agentman Studio
Call (650) 285-1019Our AI receptionist answers.