ElevenLabs Agents Review (2026)
The speech company's voice agent, priced per minute with a free tier and no sales call — and the tiers are volume commitments rather than volume discounts.
Some links on this page can be affiliate links. We may earn a commission at no cost to you. This never affects our rankings. How we work.
Visit ElevenLabs Agents ↗Opens elevenlabs.io in a new tab.
What the score measures. Five things we can verify from published material: whether pricing is transparent, whether the product can act on your systems or only answer from documents, whether the vendor still owns its own roadmap, how much independent review evidence exists, and whether the bill stays predictable as volume grows. Weighted, then scored against an ideal platform that scores 5.
BEST FOR: TEAMS THAT WANT A VOICE AGENT LIVE THIS WEEK WITHOUT A PROCUREMENT CYCLE · LAST UPDATED 2026-08-17
ElevenLabs Agents in depth
ElevenLabs is a speech company first. It became widely known for text-to-speech and voice cloning, and the agent platform is built on top of that: the differentiator it leads with is how the agent sounds rather than what it can do to your systems.
That matters for a support buyer because it sets what you are really evaluating. The synthesis quality is the best-regarded part of the product and it is the part least likely to decide whether a support deployment works. The parts that do decide it — what the agent can touch, what happens when it is unsure, what the bill looks like at real volume — are ordinary, and one of them is unusually transparent.
What it does
- Voice and chat agents deployable to phone, web chat, email and WhatsApp
- Bundled call minutes across seven tiers, with concurrency as a second axis
- Four tool types: client, webhook, MCP and built-in system tools
- Knowledge base with retrieval, plus a workflow builder
- Multilingual, with the vendor claiming 70+ languages
- BAAs for HIPAA customers on the Enterprise tier
More detail on how it works
It is also the only platform in this directory that sells a customer-facing voice agent on a free tier you can start without speaking to anyone. In a category where one vendor in nine publishes a price at all, that is the finding.
What the agent can act on
ElevenLabs documents four kinds of tool. Client tools run in your own application. Webhook tools call external APIs. MCP tools connect Model Context Protocol servers. System tools are built in for common actions. The load-bearing one for support is the webhook tool, and the documentation is direct about what it is for: it allows an assistant to “trigger authenticated actions based on the conversation, like scheduling meetings or initiating order returns”. So the agent can change something in a system of record — through an endpoint you write and authenticate. That places it firmly on the buyer-built side of the reach question. Nothing arrives wired to Shopify or Zendesk the way Gorgias or Yuma does. What you get is a capable voice layer and a documented way to hand it your API.
What the support pages claim
The customer service page leads on deflecting high-volume routine work — order tracking, billing, password resets — and names Revolut, Meesho, Deliveroo, Cisco and Deutsche Telekom among its customers. It also claims billing and account management including processing payments and plan changes. That is a strong claim for an agent whose writes run through webhooks you build, and it is worth reading as a description of what the platform permits rather than what it ships. The mechanism is yours.
Setting ElevenLabs Agents up
Realistic time to a working deployment: Weeks. These are the standard steps for this category. We have not published a walkthrough specific to this tool yet.
- 01
Scope the calls you are automating
One or two intents, not the whole queue. Voice failure is far more visible to customers than chat failure.
- 02
Provision numbers and routing
Port a number or forward from your existing system. Carrier work, and the step with actual lead times.
- 03
Build the conversation flows
Voice needs tighter scripting than chat because there is no scroll-back and no visible menu.
- 04
Connect backend systems
Order, account or appointment lookups. Same engineering work as chat, less tolerance for latency.
- 05
Tune interruption and silence handling
Barge-in, timeouts and turn-taking. Gets the least attention in evaluation and drives the most customer frustration.
- 06
Test on real accents and real noise
Not the demo audio. Your actual callers, on phones, in cars.
- 07
Keep a fast route to a human
A voice agent with no obvious escape is the fastest way to lose a customer.
The step that takes longer than they imply. Telephony. Numbers, carrier routing and IVR replacement are infrastructure changes with real lead times, not a settings page.
What the AI actually does
Every platform here says “AI agent”. It covers a bot that reads your help centre and a system that can refund a customer. This is which one you are buying.
Acts through webhook tools that call APIs you expose and authenticate.
Its own documentation names “initiating order returns” as an example, so the capability is real — but nothing ships pre-wired, which puts it on the buyer-built side of the reach split.
- Four documented tool types: client tools running in your own app, webhook tools calling external APIs, MCP tools, and built-in system tools for common actions
- The webhook documentation is explicit about the purpose: trigger authenticated actions based on the conversation, giving scheduling and order returns as examples
- MCP support is unusual in this directory and means an agent can reach anything already exposed through an MCP server rather than needing a bespoke endpoint each time
- The support pages claim billing and account management including payment processing and plan changes, which is a description of what the platform permits rather than what it ships — the endpoint is yours
- Concurrency is a published hard limit per tier, from 4 simultaneous calls to 40, which is the constraint a support queue meets first
Who Should Use It
The second list is the more useful one.
Worth a look if
You want a voice agent live this week and will not sit through a procurement cycle. It is the only customer-facing voice agent here you can start free. Voice quality is genuinely part of the brand promise and you have been disappointed by the incumbents' synthesis. You have engineering capacity to write and secure the webhooks the agent acts through.
Who should look elsewhere. Do not buy this expecting a support platform. There is no ticketing, no queue, no agent desktop and no help-desk-native integration — you are buying a voice layer that talks to systems you connect. If nobody on your side is going to build and secure those endpoints, the agent will answer questions well and change nothing.
Strengths and Weaknesses
What Works
- Publishes a complete rate card in a category that almost never does, with a free tier and a self-serve path.
- The per-minute rate is identical at every paid tier, so the bill is forecastable.
- Tooling is documented plainly across four types including MCP.
What Does Not
- LLM tokens are metered separately and vary by model, and burst calls above your concurrency limit cost double, so two of the three meters are invisible on the pricing page.
- The reach is yours to build: nothing ships wired to a help desk or a store.
- No volume discount at all below Enterprise.
- And the product's centre of gravity is speech rather than support workflow, so a lot of what a support team needs sits outside it.
ElevenLabs Agents pricing
Read from https://elevenlabs.io/pricing/agents on 2026-08-17.
$0
per month15 minutes of calls, 4 concurrent
$6
per month75 minutes, 6 concurrent
$22
per month275 minutes, 10 concurrent
$99
per month1,238 minutes, 20 concurrent
$299
per month3,738 minutes, 30 concurrent, 3 seats
$990
per month12,375 minutes, 40 concurrent, 10 seats
Custom
Elevated concurrency, custom DPA and SLA, BAAs for HIPAA customers, SSO
Verified 2026-08-17 · https://elevenlabs.io/pricing/agents · Free trial: Free tier with 15 minutes of calls, no card
ElevenAgents publishes seven tiers, all of them on bundled call minutes: Free at 15 minutes, Starter $6 for 75, Creator $22 for 275, Pro $99 for 1,238, Scale $299 for 3,738, Business $990 for 12,375, then Enterprise on request.
Divide any paid tier by its minutes and you get the same number. $6/75, $22/275, $99/1,238, $299/3,738 and $990/12,375 all come to $0.08 a minute. The tiers are volume commitments, not volume discounts — moving up buys you more included minutes and more concurrency at an identical unit rate, and the only place a discount appears is Enterprise, where the page says significant discounts at scale and stops.
The full arithmetic
That is unusually legible for this category, and it cuts both ways. You can forecast the bill exactly, and you get nothing for growing until you are big enough to negotiate.
Concurrency is the second axis and the one that bites: 4 simultaneous calls on Free, 6 on Starter, 10 on Creator, 20 on Pro, 30 on Scale, 40 on Business. A support queue with a spike does not care about your monthly minute balance if it runs out of concurrent lines.
INTEGRATIONS
Our Recommendation
Who we would tell to buy this, and who we would not.
The most transparent pricing in the voice category, on the best-regarded speech engine, attached to an agent whose ability to do anything is entirely a function of what you build behind it. That is a good trade for a team with engineers and a bad one for a support org expecting a platform. Judge it on the webhook work you are willing to do, not on the demo, because the demo is the part that was always going to be good.
Where we write about ElevenLabs Agents
3 pieces on this site name it.
- The tools that can change something are not the tools allowed near health data.Sixty platforms here can change something in a system of record. Twenty-one of them either state they do not support HIPAA or say nothing at all. Here is the overlap between what a tool can do and what data it is allowed to touch, plus a second pattern: supporting HIPAA makes a vendor half as likely to publish a price.
- Every voice platform here can change something in your systems. Two of them will tell you what that costs.Voice is the only category in this directory where every platform can act on a system of record, and it is near the bottom on whether anyone will quote you. Here is what the two that publish actually charge, the meters that sit underneath a per-minute rate, and why a quoted price per minute is not comparable across vendors.
- Every paid ElevenLabs tier costs exactly the same per minute. The interesting part is what the tier table leaves out.Divide any paid ElevenAgents tier by its included minutes and you get $0.0800. Starter, Creator, Pro, Scale, Business — identical. The tiers are volume commitments rather than volume discounts. Here is the arithmetic, the two meters the pricing page does not show, and the one clause that goes in your favour.
Similar Tools
Others in voice & phone ai.
Frequently Asked
The questions people actually search for about ElevenLabs Agents.
How much does an ElevenLabs voice agent cost?
Every paid tier works out at exactly $0.08 a minute. Starter is $6 for 75 minutes, Creator $22 for 275, Pro $99 for 1,238, Scale $299 for 3,738 and Business $990 for 12,375. There is a free tier with 15 minutes. Enterprise is quoted, and is the only place the page mentions a discount.
Is there a free tier?
Yes — 15 minutes of calls and 4 concurrent calls, with the workflow builder, knowledge base and multilingual widget included. It is the only customer-facing voice agent in this directory you can start without talking to sales.
Can it actually do things, or only talk?
It can act, through webhook tools that call APIs you expose and authenticate. Its own documentation gives initiating order returns as an example. But the integration is yours to build — nothing arrives pre-wired to a help desk or store, which is the opposite of how Gorgias or Yuma ship.
Is it HIPAA compliant?
BAAs for HIPAA customers are listed on the Enterprise tier only, alongside custom DPA and SLA terms. Nothing below Enterprise offers one, so a regulated buyer is in a sales conversation regardless of the self-serve pricing above it.
What is the catch with the pricing?
The $0.08 is one of three meters. LLM inference is billed separately by input and output tokens and varies with the model you pick, and burst calls above your concurrency limit are charged at double the standard rate. Neither appears in the tier table. There is also no volume discount below Enterprise. The one thing that goes your way is silence, which bills at 5% of the normal per-minute rate.
Is this a replacement for a help desk?
No. There is no ticketing, queue or agent desktop. It is a voice and chat layer that connects to the systems you already run, and it is priced and documented as one.