ElevenLabs integrations: n8n, Zapier, Twilio and what needs a custom build

This page may contain affiliate links from which Blur Studio earns commissions.

ElevenLabs, the AI voice platform for text to speech, voice cloning and conversational agents, connects to other tools in a few genuinely different ways, some built and maintained directly by ElevenLabs, some through third-party automation platforms, and some that realistically need custom integration work. This guide covers connection options, three setups that actually work as described by the platforms themselves, and what's honestly a custom-build job rather than a flip of a switch.

Connection options

• n8n, a workflow automation platform for connecting apps and APIs together without writing custom code: an official integration, described on n8n's own site as "built and maintained by our partners at ElevenLabs and verified by n8n." It covers text to speech, speech to text, speech to speech (converting audio from one voice to another), and voice management (creating clones, fetching voice metadata), plus a custom API call option for anything not covered by the built-in actions.
• Zapier, another workflow automation platform that links actions across different apps: a listed integration centered on a Convert Text to Speech action, which takes text and a voice selection and returns generated audio inside a Zap. Zapier's own page also lists MCP, SDK and CLI as additional ways to connect, beyond the standard Zap-building interface.
• Twilio, a cloud telephony provider that supplies phone numbers and call infrastructure: a native integration specifically for ElevenLabs Agents, letting you import a Twilio phone number directly into the ElevenLabs dashboard for both inbound call routing to an agent and outbound calling, with the platform handling the number configuration automatically.
• MCP (Model Context Protocol): referenced by Zapier as a way to expose ElevenLabs operations as tools to any MCP-compatible AI client, relevant if you're building with an AI assistant or agent framework that supports connecting external tools this way.
• WhatsApp: I could not confirm a dedicated, native ElevenLabs-to-WhatsApp integration directly from ElevenLabs' own documentation during this research, several relevant pages returned access-blocked responses. If you need agent conversations over WhatsApp specifically, treat this as something to build through a webhook or a provider that bridges WhatsApp's Business API, and confirm the current, correct setup path directly with ElevenLabs before assuming a built-in option exists.

Three working setups

1. Automated voice alerts through n8n: trigger a workflow off an event (a new support ticket, a form submission, a monitoring alert) and have n8n call ElevenLabs' text to speech action to generate a spoken alert, then route that audio wherever it needs to go, a phone call, a Slack message, a stored file. This uses only the official, maintained n8n node, no custom code required. Pushing call data into Google Sheets or Slack is a related no-code setup.
2. Generated narration through Zapier: connect a content source (a new row in a spreadsheet, a new document, a CMS entry) to Zapier's Convert Text to Speech action, so new written content automatically gets a voiced version generated without manual steps, useful for a steady stream of short-form content that needs a consistent voice.
3. A phone agent on a real number through Twilio: import an existing or new Twilio number directly into the ElevenLabs dashboard, assign it to a configured agent, and inbound calls route straight to that agent, covered in more depth in my ElevenLabs Agents guide.

What needs custom build

• WhatsApp conversations with an agent, since there's no confirmed native toggle for this, it needs a webhook bridge or a WhatsApp Business API provider connected to your agent through the API.
• Deep MCP integration into a custom AI tool or internal assistant, beyond the basic tool-exposure Zapier references, generally needs a developer to wire up properly for a specific internal use case.
• Anything requiring data from an internal system (a CRM, a billing platform, an inventory system) feeding into an agent's knowledge base or a TTS automation in real time, rather than a one-off manual upload, usually needs custom API work beyond what n8n or Zapier's pre-built actions cover. Letting an AI phone agent look up an order or ticket status is one example.

A reasonable rule of thumb before deciding which path to take: if the integration is a single, repeatable step (generate audio from new text, transcribe a new recording), start with n8n or Zapier, since the pre-built action already exists and needs no custom code. If the integration involves several conditional steps, data pulled live from a system that doesn't have a pre-built connector, or a channel without a confirmed native option, plan for custom API work from the start rather than trying to force an automation platform's action into a shape it wasn't built for. Letting an AI voice agent book jobs into your calendar or CRM shows the custom side.

Choosing between n8n and Zapier

Both platforms can trigger an ElevenLabs text to speech generation from an automated workflow, but they fit different teams. n8n's ElevenLabs integration is the deeper of the two, covering text to speech, speech to text, speech to speech and voice management in one node, which matters if your workflow needs more than a single text-to-audio step, converting a recorded call to text and then generating a different audio output from it, for example. n8n also supports self-hosting, which some businesses prefer for data control, while Zapier is exclusively a hosted service. Zapier's strength is the sheer number of other apps it already connects to out of the box, over 30 popular combinations are highlighted directly on ElevenLabs' Zapier listing alone, Google Drive, Gmail, Notion, Airtable, HubSpot, Salesforce and more, which matters more if your priority is connecting ElevenLabs into an existing stack of mainstream business tools rather than building a more involved multi-step automation.

If you're not sure which to start with, Zapier is generally the faster path to a first working automation if you already use several of the apps it lists prominently. n8n is the better choice once you need more than one ElevenLabs action chained together in a single workflow, or you want the option to self-host.

A note on webhooks

Separate from both pre-built automation platforms, ElevenLabs' API supports webhooks directly, letting an event on ElevenLabs' side (a completed dubbing job, for instance) trigger a callback to your own system rather than your system having to repeatedly check whether a job finished. This is the lower-level building block that n8n, Zapier and any custom integration ultimately sit on top of, and it's the right starting point if neither pre-built platform's actions cover your specific need and you're building a custom connection yourself or with a developer.

A quick sanity check before you build

Before setting up any integration, confirm the specific action you need actually exists in the platform you're planning to use, rather than assuming it does because the two products are listed as "integrated." n8n's ElevenLabs node, for example, covers several distinct actions, but not necessarily every single API capability ElevenLabs offers, newer features sometimes land in the raw API before a pre-built automation action catches up. A five-minute check of the actual action list in your automation platform's own documentation before you commit to a workflow design saves the frustration of building most of an automation around a step that turns out not to exist yet.

How I'd approach a custom integration

When a client needs ElevenLabs connected to something n8n and Zapier's built-in actions don't cover, a specific internal system, a WhatsApp flow, a custom MCP tool setup, I build it directly against the ElevenLabs API rather than forcing an automation platform to do something it wasn't built for. Get in touch if you want an integration built properly around your actual systems instead of patched together.