Realtime Voice Agent APIs
Speech-to-speech and realtime conversation APIs for building voice agents over phone or web.
TL;DR: Top pick: ElevenLabs Conversational AI (ElevenAgents). 14 APIs compared, 5 with a free tier; cheapest published entry $0.01 minute (Pipecat Cloud (Daily)). Why →
What is a Realtime Voice Agent API?
A Realtime Voice Agent API lets developers speech-to-speech and realtime conversation apis for building voice agents over phone or web over HTTP. The Realtime Voice Agent APIs below are compared on pricing, compliance, capabilities, and developer experience from structured, cited data.
| # | API | Pricing | Free | Self-serve | SOC 2 | HIPAA | GDPR | Webhooks | Actions |
|---|---|---|---|---|---|---|---|---|---|
| 1 | ElevenLabs Conversational AI (ElevenAgents) | Hybrid | ✓ | ✓ | Type II | ✓ | ✓ | ✓ | 18 |
| 2 | OpenAI Realtime API (gpt-realtime) | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 17 |
| 3 | Telnyx Voice AI Agents | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 26 |
| 4 | Twilio ConversationRelay | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 23 |
| 5 | Vapi | Usage | ✓ | ✓ | Type II | ✓ | ✓ | ✓ | 26 |
| 6 | Cartesia Line | Hybrid | ✓ | ✓ | Type II | ✓ | ✓ | ✓ | 25 |
| 7 | Pipecat Cloud (Daily) | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 23 |
| 8 | LiveKit Agents | Hybrid | ✓ | ✓ | Type II | ✓ | ✓ | ✓ | 22 |
| 9 | Bland AI | Hybrid | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 22 |
| 10 | Deepgram Voice Agent API | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✗ | 16 |
| 11 | Retell AI | Usage | ✗ | ✓ | Type II | ✓ | ✓ | ✓ | 29 |
| 12 | Synthflow AI | Usage | ✗ | ✓ | ? | ✓ | ✓ | ✓ | 23 |
| 13 | Ultravox | Usage | ✗ | ✓ | · | · | · | ✓ | 20 |
| 14 | Vocode | Usage | ✓ | ✓ | · | ✗ | · | ✓ | 19 |
The APIs
#1 ElevenLabs Conversational AI (ElevenAgents)
81 / 100ElevenLabs Conversational AI (ElevenAgents) is a real-time voice agent platform for building inbound and outbound phone automation, covering use cases from customer support and appointment scheduling to healthcare answering services and sales calls. Pricing starts at $0.08 per minute with a free tier of 15 minutes per month, scaling through self-serve plans up to enterprise; telephony costs are billed separately. The platform holds SOC 2 Type 2, HIPAA, GDPR, ISO 27001, and PCI DSS certifications, with SDKs for Python, TypeScript, React, React Native, and Kotlin, plus an MCP server and support for bring-your-own LLM and voice.
PricingHybrid · from $0.08 minute · free tier ✓TrustSOC 2 Type II · HIPAA · GDPR · ISO 27001 · PCI DSSDoesUsed byRevolut, Deliveroo, Epic Games, Deutsche Telekom#2 OpenAI Realtime API (gpt-realtime)
72 / 100OpenAI Realtime API is a WebSocket-based service for low-latency, bidirectional speech-to-speech communication, targeting developers building voice agents, real-time translation, live transcription, and call center automation. Pricing is usage-based starting at $0.0192 per minute of audio input, with self-serve signup and no sales call required. The API supports function calling, voice activity detection, interruption handling, and inbound SIP telephony via a third-party carrier. It holds SOC 2 Type II, HIPAA, GDPR, ISO 27001, and PCI DSS certifications, with SDKs available for Python, Node.js, Go, Java, Ruby, and .NET.
PricingUsage · from $0.02 minute (audio input) · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · ISO 27001 · PCI DSSDoesUsed byPerplexity, Healthify, Speak, ZillowAvoid ifYou want to try it free before paying#3 Telnyx Voice AI Agents
78 / 100Telnyx Voice AI Agents is a carrier-grade platform for building inbound and outbound voice AI, covering customer support automation, appointment scheduling, healthcare, and call center use cases across 45+ countries. Pricing is usage-based at $0.05 per minute with self-serve signup and no sales call required. The platform supports bring-your-own LLM and voice, function calling, SIP trunking, and SDKs for Node.js, Python, PHP, Java, Ruby, and Go. It holds SOC 2 Type II, HIPAA, GDPR, ISO 27001, and PCI DSS certifications, and customers include Dialpad, Alibaba, and Grupo Bimbo.
PricingUsage · from $0.05 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · ISO 27001 · PCI DSSDoesUsed byDialpad, Alibaba, AudioCodes, UJETAvoid ifYou want to try it free before paying#4 Twilio ConversationRelay
76 / 100Twilio ConversationRelay is a WebSocket-based orchestration layer that handles speech recognition, text-to-speech, and real-time conversation flow for AI voice agents over PSTN, SIP, and WebRTC, supporting bring-your-own LLM and voice providers. It targets developers building inbound and outbound call automation, such as customer support, appointment scheduling, and live agent escalation. Pricing starts at $0.07 per minute for the orchestration layer, billed separately from underlying voice costs, with self-serve signup and SDKs for seven languages. The service reached general availability in May 2025 and holds SOC 2 Type II, HIPAA, GDPR, ISO 27001, and PCI DSS certifications, though HIPAA coverage requires a Security or Enterprise Edition account.
PricingUsage · from $0.07 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · ISO 27001 · PCI DSSDoesUsed byOhMD, ScorpionAvoid ifYou want to try it free before paying#5 Vapi
75 / 100Vapi is a voice agent platform for building and deploying AI-powered phone agents handling inbound calls, outbound dialing, appointment scheduling, and lead qualification. Pricing is usage-based at $0.05 per minute with self-serve signup and no sales call required, plus an enterprise plan for custom concurrency. The REST API supports bring-your-own LLM and voice providers, ships SDKs for over ten languages and runtimes including mobile, and holds SOC 2 Type 2, HIPAA, GDPR, and PCI DSS certifications, with customers including New York Life and Intuit.
PricingUsage · from $0.05 minute · free tier ✓TrustSOC 2 Type II · HIPAA · GDPR · PCI DSSDoesUsed byAmazon Ring, Kavak, Instawork, New York Life#6 Cartesia Line
79 / 100Cartesia Line is a telephony platform for building and deploying AI voice agents, supporting inbound call handling, outbound calling, batch dialing, SIP trunking, and real-time conversation via a WebSocket API. It targets developers and businesses automating customer support or voice workflows, with customers including ServiceNow, Vapi, LiveKit, and Replicant. Pricing starts at $0.06 per minute with a free tier of 8 concurrent calls and 20,000 credits per month; enterprise plans are available. The platform holds SOC 2 Type 2, HIPAA, GDPR, and PCI DSS certifications, and ships SDKs for Python and JavaScript/TypeScript.
PricingHybrid · from $0.06 minute · free tier ✓TrustSOC 2 Type II · HIPAA · GDPR · PCI DSSDoesUsed byServiceNow, Maven AGI, Retell, Vapi#7 Pipecat Cloud (Daily)
64 / 100Pipecat Cloud is a managed hosting platform from Daily for deploying and scaling Pipecat voice AI agents in production, supporting inbound and outbound telephony via PSTN/SIP, WebRTC, and WhatsApp. It targets teams building customer support bots, IVR systems, and outbound call automation, with self-serve signup and usage-based pricing starting at $0.01 per minute. The platform spans four regions, offers Python and JavaScript SDKs with MCP server support, and holds SOC 2 Type 2, ISO 27001, HIPAA, and GDPR certifications. Notable users include NVIDIA, Descript, and Epic.
PricingUsage · from $0.01 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · ISO 27001DoesUsed byNVIDIA, Mercor, Descript, EpicAvoid ifYou want to try it free before paying#8 LiveKit Agents
71 / 100LiveKit Agents is a developer platform for building and deploying realtime voice and video AI agents, supporting inbound and outbound telephony, SIP trunking, multi-agent orchestration, and bring-your-own LLM, STT, and TTS integrations. Pricing starts at $0.01 per minute with a free tier of 1,000 agent session minutes per month and no credit card required, scaling through tiered plans up to custom enterprise arrangements. SDKs are available for Python, Node.js, and Go, and the platform holds SOC 2 Type 2, HIPAA, and GDPR certifications, with HIPAA compliance restricted to Scale and Enterprise plans. Customers include OpenAI, Salesforce, Oracle, and Deutsche Telekom.
PricingHybrid · from $0.01 minute · free tier ✓TrustSOC 2 Type II · HIPAA · GDPRDoesUsed byOpenAI, Oracle, Salesforce, Deutsche Telekom#9 Bland AI
63 / 100Bland AI is a REST-based voice AI platform for automating inbound and outbound phone calls, targeting regulated industries such as insurance, healthcare, and finance, with customers including Mutual of Omaha and First Financial Bank. Pricing starts at $0.11 per minute on a self-serve plan, with tiered plans scaling up to enterprise unlimited concurrency. The platform holds SOC 2 Type 2, HIPAA, GDPR, and PCI DSS certifications, and supports EU data residency and air-gapped on-premises deployment for organizations with strict data requirements.
PricingHybrid · from $0.11 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · PCI DSSDoesUsed byKin Insurance, Mutual of Omaha, TravelPerk, SamsaraAvoid ifYou want to try it free before paying#10 Deepgram Voice Agent API
68 / 100Deepgram Voice Agent API is a single WebSocket-based interface for building real-time voice AI agents, covering inbound and outbound telephony, function calling, interruption handling, and multi-agent orchestration, with SDKs for Python, Node.js, Go, and C#/.NET. Pricing is usage-based at $0.075 per minute with a one-time $200 trial credit and self-serve signup, scaling to enterprise plans with 100 or more concurrent connections. The platform holds SOC 2 Type 2, HIPAA, GDPR, and PCI DSS certifications and reached general availability in June 2025, with customers including NASA, IBM, Five9, and Aircall. Telephony requires Twilio or Amazon Connect for PSTN access, and the Growth plan carries a $4,000 per year minimum commitment.
PricingUsage · from $0.07 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPR · PCI DSSDoesUsed byAircall, Jack in the Box, OpenPhone, NASAAvoid ifYou want to try it free before paying#11 Retell AI
68 / 100Retell AI is a voice agent platform for automating phone calls, supporting both inbound and outbound use cases such as customer support, appointment booking, lead qualification, and call center operations. Pricing is usage-based at $0.055 per minute with a $10 free credit on signup, and an enterprise plan is available for higher-volume needs. The REST API supports bring-your-own LLM and voice, SIP trunking, WebRTC web calling, and function calling, with SDKs for Python, Node.js, and browser JavaScript. The platform holds SOC 2 Type II certification and is HIPAA and GDPR compliant, though it currently operates only from US-based infrastructure.
PricingUsage · from $0.06 minute · free tier ✗TrustSOC 2 Type II · HIPAA · GDPRDoesUsed byLenovo, Asbury Auto, Anker, StorageVaultAvoid ifYou want to try it free before paying#12 Synthflow AI
57 / 100Synthflow AI is a voice AI platform for automating inbound and outbound phone calls, targeting use cases such as appointment booking, lead qualification, customer support, and healthcare reception. Pricing is usage-based at $0.09 per minute with self-serve signup, no sales call required, and an enterprise plan available. The REST API supports webhooks, warm call transfers, voicemail detection, and sentiment analysis, with HIPAA, GDPR, ISO 27001, and PCI DSS compliance documented. Default PAYG accounts are capped at five concurrent calls, expandable for an additional fee.
PricingUsage · from $0.09 minute · free tier ✗TrustHIPAA · GDPR · ISO 27001 · PCI DSSDoesUsed byFreshworksAvoid ifYou want to try it free before paying#13 Ultravox
44 / 100Ultravox is a speech-native voice AI API for building real-time conversational agents, targeting developers who need inbound and outbound telephony automation, AI receptionists, or voice-enabled web and mobile apps. Pricing is usage-based at $0.05 per minute with 30 free minutes included and self-serve signup, though pay-as-you-go accounts are hard-capped at 5 concurrent calls and the outbound call scheduler requires a Pro plan or higher. The platform integrates with Twilio, Telnyx, Plivo, and jambonz for telephony but does not provision phone numbers itself. SDKs are available for JavaScript, Python, React Native, Flutter, Android, and iOS.
PricingUsage · from $0.05 minute · free tier ✗DoesUsed by11xAvoid ifYou want to try it free before paying#14 Vocode
53 / 100Vocode is an MIT-licensed open-source library for building real-time voice agents, covering inbound and outbound telephony, IVR navigation, call transfer, answering machine detection, and function calling. It targets developers building AI phone agents for call centers, customer support, appointment booking, and sales. Paid plans start at $25 per seat per month with a free tier included, though the hosted API appears discontinued and active development has stalled since mid-2024. SDKs are available for Python and Node.js, with REST API access authenticated via API key.
PricingUsage · free tier ✓DoesAvoid ifYou need transparent pricing up front