ElevenAgents — Independent Software Review

Deploy AI Agents in Minutes, Not Months

Compliance Transparency Index

Grade: A — Score: 88/100

Best For

• Customer-support teams adding voice alongside web or app chat. Workflows can route specialist tasks and escalate to a human phone number.
• Sales and booking operations connecting conversations to business records. Documented Salesforce and Google Calendar integrations support CRM lookups and scheduling actions.
• Multilingual teams prepared to test each target language and voice. Eleven v3 Conversational supports 70+ languages with prompt-guided delivery and configurable turn-taking.
• Engineering teams wanting managed orchestration with API access. JSON workflows, SDK deployment and three types of agent tests support controlled releases.
• Enterprises able to negotiate deployment and data-processing requirements. SSO, a BAA route, retention controls and private deployment are documented, with contractual and configuration conditions.

Not Ideal For

• Support buyers needing charges tied to resolved outcomes rather than conversation activity. Call duration, text, LLM and telephony are separately metered, so the advertised subscription is not an all-inclusive operating budget.
• Small teams expecting high peak concurrency on a low subscription. Free includes 4 simultaneous calls and Starter 6; burst capacity doubles the platform rate and receives lower-priority speech processing.
• Builders needing an openly available self-hosted runtime or unrestricted, ready-made data connectors. Private deployment documentation is sales/NDA-gated, native HubSpot excludes EU-hosted accounts, and document uploads and indexed knowledge have explicit size limits.
• Regulated teams needing one unambiguous public HIPAA retention instruction. The BAA route is Enterprise-only, persisted agent configuration is excluded, MCP cannot be used in HIPAA or Zero Retention Mode workspaces, and the general audio-retention recommendation conflicts with the HIPAA-specific Zero Retention requirement.
• Brand-voice teams requiring v3 Conversational expressiveness and faithful reproduction of an existing Professional Voice Clone at the same time. The vendor documents that v3 does not preserve those clone characteristics well and recommends Flash v2 when identity is critical.

Operational Overview

Pricing Structure

Free: USD $0/month; permanent free plan, not a paid-tier trial

• 15 call minutes per month; 4 concurrent calls
• Workflow Builder, knowledge base and web widget
• 1 MB of documents indexed for RAG per workspace
• General terms restrict free use to non-commercial purposes; the landing-page start card does not scope its adjacent commercial-licence wording to a paid plan

Starter: USD $6/month, billed monthly; or $5/month equivalent, billed yearly. Taxes extra

• 75 call minutes per month; 6 concurrent calls
• 2 MB of documents indexed for RAG per workspace
• Paid commercial-use permission, subject to the applicable terms

Creator: USD $22/month, billed monthly; or $18.33/month equivalent, billed yearly. Taxes extra

• 275 call minutes per month; 10 concurrent calls
• 20 MB of documents indexed for RAG per workspace

Pro: USD $99/month, billed monthly; or $82.50/month equivalent, billed yearly. Taxes extra

• 1,238 call minutes per month; 20 concurrent calls
• 100 MB of documents indexed for RAG per workspace

Scale: USD $299/month, billed monthly; or $249.17/month equivalent, billed yearly. Taxes extra

• 3,738 call minutes per month; 30 concurrent calls
• 3 workspace seats
• 500 MB of documents indexed for RAG per workspace

Business: USD $990/month, billed monthly; or $825/month equivalent, billed yearly. Taxes extra

• 12,375 call minutes per month; 40 concurrent calls
• 10 workspace seats and 10 Professional Voice Clones
• 1 GB of documents indexed for RAG per workspace
• Clone count does not remove the v3 Conversational voice-fidelity limitation

Enterprise: Custom quote; prices, minimum commitments and negotiated terms are not public

• Negotiated concurrency, seats and support
• SSO and custom contractual assurances
• BAA available for eligible HIPAA deployments with Zero Retention Mode
• Higher RAG limits by agreement; published baseline is 1 GB

Pay As You Go and additional usage: Platform charges: USD $0.080/additional call minute, $0.003/text message and $0.160/burst call minute; LLM usage and telephony extra

• Prepaid top-ups from USD $5 or INR 500; available on self-service plans including Free
• Model-dependent LLM charges are separate; telephony is passed through at cost
• Top-ups extend usage, not plan permissions or normal concurrency
• Optional automatic top-ups; prepaid balance expires 12 months after purchase
• Legacy usage-based subscriptions have a separate migration path

Alternative Consideration

Consider switching to Google Dialogflow: Google Dialogflow offers a well-established platform with extensive integration options and a large developer community.

Frequently Asked Questions

How does ElevenAgents compare with Retell AI for a phone agent?

Retell AI includes 20 concurrent calls on pay-as-you-go, while ElevenAgents includes 4 on Free, 6 on Starter and 20 on Pro before optional burst capacity. ElevenAgents lists additional platform minutes at USD $0.080 plus LLM and telephony charges; Retell advertises USD $0.07-$0.31 per voice minute, with configuration-dependent telephony and add-ons, so these are not directly comparable all-in rates. Retell sells additional concurrency at USD $8 per simultaneous call per month, whereas ElevenAgents burst calls incur twice the standard platform rate and receive lower-priority speech processing. Compare both the complete call configuration and peak traffic, not just the headline minute price.

How does ElevenAgents compare with LiveKit Agents if I want to self-host?

ElevenAgents provides a managed visual workflow editor, while LiveKit Agents offers an Apache 2.0 framework for building agents in Python or Node.js with selectable speech and language-model providers. ElevenAgents also offers its full runtime in your own cloud or hardware, but detailed private-deployment documentation requires authorization and an NDA for prospective customers. LiveKit supports self-hosted agent servers as well as a separate managed Cloud option, which also has a browser-based Agent Builder. Self-hosting LiveKit means taking responsibility for infrastructure, integration code, scaling and provider charges; an open-source framework does not make inference free or make an ElevenLabs voice clone portable.

Can ElevenAgents use my existing business phone number without moving everything to Twilio?

ElevenAgents documents SIP trunking for connecting an existing PBX or SIP-enabled phone system and handling inbound and outbound calls. Its native Twilio integration is another route: purchased Twilio numbers support both directions, but verified caller IDs support outbound calls only. Verifying an existing number as a Twilio caller ID therefore does not make that number an inbound ElevenAgents line. Check your carrier's SIP compatibility, routing and authentication requirements before choosing the connection method.

Can ElevenAgents transfer a caller to a person with a warm handoff?

ElevenAgents supports human handoffs through configurable transfer rules for external phone numbers and, with SIP REFER, SIP addresses. A conference transfer can play a briefing message to the human recipient when the agent uses the native Twilio integration. SIP-based transfers do not support that warm-transfer message, and SIP REFER requires a trunk that permits it. Do not assume a configured transfer guarantees an answered call; test busy, no-answer and timeout behavior with your carrier before deployment.

Can ElevenAgents book appointments and update my CRM, rather than just discuss them?

ElevenAgents documents a Google Calendar integration that checks availability and creates events through authorized tool calls, plus a native HubSpot integration that searches, creates and updates CRM records. You must connect the accounts, grant appropriate permissions and configure when the agent should invoke each action. The native HubSpot connector currently supports US-hosted accounts only, not EU-hosted accounts. Use Tool Call tests to check invocation and parameters, then test the complete booking workflow and its failure cases; a spoken confirmation alone is not evidence that an appointment was saved.

Can ElevenAgents answer accurately from a large knowledge base?

ElevenAgents supports documents, URLs and text, with retrieval-augmented generation selecting relevant source material rather than placing every document in the prompt. There is a 20 MB upload limit per file and a separate workspace-wide allowance for indexed original documents, such as 1 MB on Free and 100 MB on Pro. Its testing framework includes full conversation simulations, next-reply checks and tool-call tests that can be used to evaluate your configured agent. Retrieved material and passing tests are not guarantees of factual accuracy, so check representative questions and cases where the source does not contain an answer before deployment.

Can ElevenAgents use my Professional Voice Clone with v3 Conversational expressiveness?

Eleven v3 Conversational supports expressive delivery in 70+ languages, but ElevenLabs explicitly warns that it does not preserve Professional Voice Clone characteristics well. The resulting voice may therefore differ from your existing clone even when that clone is selected. If preserving the clone's identity is essential, the vendor recommends considering Flash v2 instead. Treat expressive delivery and faithful clone reproduction as separate requirements, and test the chosen model in each target language.

What does ElevenAgents require before a healthcare agent can handle PHI?

ElevenAgents is HIPAA-eligible subject to an executed Enterprise Business Associate Agreement, enabled Zero Retention Mode and an eligible language-model configuration. The general privacy page recommends saving audio for at least six years for HIPAA, while the HIPAA-specific guide requires Zero Retention Mode and says PHI audio and transcripts are not retained. Zero Retention Mode covers eligible API traffic rather than web-interface or playground traffic, persisted agent configuration is excluded from PHI coverage, and MCP integrations are unavailable. Obtain a reconciled instruction from ElevenLabs and review the agreement and complete data path before sending protected health information.

Can I stop ElevenAgents data from being used to train ElevenLabs models?

ElevenLabs documents a training opt-out under Terms and privacy, then Data use: disable Improve the models for everyone and save the choice. The stated protection applies to newly submitted data, not a reversal of earlier training. Enterprise data is excluded from training by default except as necessary to provide that customer's services. Training opt-out is separate from conversation retention and Zero Retention Mode, so configure those controls independently.

Can I export and delete ElevenAgents call transcripts and recordings?

ElevenAgents provides APIs to retrieve retained conversation transcripts as JSON, download available call audio and delete individual conversations. Conversation data is retained for two years by default, with separately configurable transcript and audio retention periods. With Zero Retention Mode enabled, the vendor directs you to post-call webhooks to collect call content you need in your own systems rather than relying on stored call history. Export before deletion or expiry, and manage downstream copies separately; the documented conversation API does not establish deletion of every copy held by your other systems.

AI Visibility Report

How AI agents (ChatGPT, Perplexity, Claude, others) read this review page in the past 7 days. Updated weekly. View ElevenAgents AI Visibility Report.