LiveKit pricing starts at $0 with the Build plan. I've tested all four tiers, and here's exactly what each one costs, what's included, and which plan makes sense for your voice AI project. ## LiveKit Pricing Plans: At a Glance | 🏆 Plan | 💰 Price | 🎯 Best For | ⚡ Key Features | |---------|---------|------------|----------------| | **Build** | $0/month | First projects and local development | 1,000 agent session minutes; $2.50 in inference credits; 1 free US number; community support | | **Ship** | Starting at $50/month | Shipping to production users | 5,000 agent session minutes; 20 custom voices; rollback to previous deployments; email support | | **Scale** | Starting at $500/month | Scaling with security and global reach | 50,000 agent session minutes; 50 custom voices; inference discounts; HIPAA/SOC 2; region pinning | | **Enterprise** | Custom | High-volume teams needing SLA and compliance | Volume inference pricing; SSO; shared Slack channel; support SLA; dedicated solutions engineer | ## LiveKit Pricing Plans Breakdown Each [LiveKit plan](https://livekit.com/pricing) charges differently depending on **how much your application runs** and what production guarantees you need. ### Build: $0/month **What's included:** The Build plan covers the essentials for testing a voice agent end-to-end. You get 1,000 agent session minutes, $2.50 in inference credits, and one free US local phone number. Telephony, noise suppression, voice isolation, and observability **are all in the plan**, though at lower volumes than paid tiers. Every allowance works as a hard cap. When a limit is reached, service stops and requests fail until the next billing cycle. Agents deployed on build **may also shut down between sessions**, which can cause cold starts of up to 10 to 20 seconds. **Best for:** Solo developers testing a voice agent concept or teams evaluating LiveKit before any production usage. **Pros:**

✅ No credit card required. The plan is free with no trial countdown.

✅ Telephony, inference, observability, and noise suppression are all included out of the box, enough to validate a full concept.

**Cons:**

❌ Hard caps with no overage option mean service stops when limits are reached, which makes this plan unsuitable for anything touching production users.

❌ Cold starts on deployed agents add up to 10 to 20 seconds of delay, which users notice immediately on a voice call.

### Ship: Starting at $50/month **What's included:** Ship takes everything from build and removes the hard caps, converting each limit into a soft threshold with per-unit overages. Agent **session minutes increase** to 5,000 per month, inference credits double to $5, and WebRTC minutes jump to 150,000. The plan also opens up 20 custom voices per project, instant rollback, and cold start prevention so agents stay warm and ready. Team collaboration and email support **come with the plan** as well. **Best for:** Teams shipping a voice agent to production users for the first time, or early-stage startups validating product-market fit in production. **Pros:**

✅ Overages are allowed, so production deployments keep running when traffic spikes beyond the included allowance.

✅ Custom voices, instant rollback, and cold start prevention are all available here, which is usually what teams need before anything else.

**Cons:**

❌ Agent session overages at $0.01 per minute add up faster than expected. Ten thousand extra minutes beyond the included allowance costs $100 on top of the base fee.

❌ No HIPAA or SOC 2 compliance on this tier, which blocks any team operating in healthcare, fintech, or legal from using this plan in production.

### Scale: Starting at $500/month **What's included:** Scale is where LiveKit pricing moves from per-project cost to unit economics. The plan includes 50,000 agent session minutes and $50 in inference credits. **Model rates drop** at this tier too. ElevenLabs Flash TTS [falls by 60%](https://livekit.com/pricing/inference) compared to Ship, and Cartesia Sonic 3 TTS drops by 25%. The compliance layer activates here, covering [HIPAA and SOC 2 Type II](https://livekit.io/security) without a separate add-on. Region pinning, role-based access, and up to 600 **concurrent agent sessions** complete the tier. **Best for:** Production applications with **consistent high-volume usage**, teams in regulated industries, and deployments where inference cost directly affects unit economics. **Pros:**

✅ Inference discounts at Scale meaningfully reduce cost per conversation at volume, particularly on TTS, where ElevenLabs Flash drops 60%, and Cartesia Sonic 3 drops 25%.

✅ HIPAA and SOC 2 are included at no extra compliance cost, which matters for healthcare and fintech teams where compliance is a hard requirement before go-live.

**Cons:**

❌ The $500/month floor is a firm commitment, and teams with unpredictable or seasonal traffic will overpay in months where volume drops below the included allowances.

❌ SSO and a shared Slack support channel require upgrading to Enterprise, and there's no intermediate tier for teams that need those features before committing to a custom contract.

### Enterprise: Custom pricing **What's included:** Enterprise includes everything in Scale and replaces all published limits with custom allowances provisioned around your actual usage. Inference pricing **moves from fixed discounted rates** to negotiated volume pricing, which changes the per-conversation cost at tens of millions of minutes per month. The tier adds SSO, a shared Slack channel with the LiveKit team, a designated solutions engineer, a contractual support SLA, and non-credit **card billing via invoice** or wire transfer. **Best for:** Large teams running tens of millions of agent session minutes per month, organizations that need a formal SLA, and companies with strict SSO or data residency requirements. **Pros:**

✅ Volume inference pricing lowers the per-conversation cost once Scale's published rates become the constraint.

✅ Custom limits are provisioned in advance, so there are no platform-imposed ceilings on concurrency or session volume.

**Cons:**

❌ No published pricing means the conversation starts with a sales call before you know the number. Teams with early-stage or unpredictable usage patterns take on financial exposure before the contract makes economic sense.

## Which LiveKit Pricing Plan Should You Choose? **Choose Build if you:** - You're still in development with no production users yet - You want to validate your agent's STT, TTS, and LLM pipeline before paying anything - You can work within hard monthly caps that stop service when exceeded **Choose Ship if you:** - You're taking a voice agent to production and can't have traffic spikes cutting off active sessions - Cold start prevention and custom voices matter for the experience you're building - Your monthly volume doesn't yet justify a $500/month floor **Choose Scale if you:** - Your production traffic has stabilized enough that $500/month makes more sense than Ship's overage costs - You're in a regulated industry where HIPAA and SOC 2 Type II are requirements - You need region pinning, role-based access, or more than 20 concurrent agent sessions **Choose Enterprise if you:** - You're past tens of millions of agent session minutes per month, and volume inference pricing lowers your per-conversation cost - A formal support SLA, SSO, or a dedicated solutions engineer is part of the deal - Scale's published limits don't fit your deployment, and you need custom capacity provisioned in advance ## Is LiveKit Worth the Cost? LiveKit pricing consolidates what would otherwise be five vendor relationships into one bill. Agent deployment, inference, telephony, observability, and WebRTC **transport all land on the same invoice**. If you've priced out those layers separately, you may well find that the consolidated bill lands lower than expected, mostly because LiveKit's inference **rates are competitive** with direct vendor pricing once you're past the free tier. **LiveKit is worth it if you:** - You're building a production voice agent and want one bill instead of managing separate vendors for inference, telephony, and media transport - You're on Scale, and the inference discounts partially offset the base cost at high monthly volumes - HIPAA or SOC 2 compliance is a requirement, and Scale includes both without a separate vendor contract - Cold start prevention and instant rollback aren't optional for your product **Skip LiveKit if you:** - Your traffic is unpredictable or seasonal, and the $500/month Scale floor will hurt in slow months - You already have STT, TTS, and telephony vendor contracts at negotiated rates that undercut LiveKit Inference pricing - You're at an early stage, and Ship's per-minute overage model creates billing exposure with no ceiling to protect you ## LiveKit Pricing Alternatives & Comparison | 🛠 Tool | 💰 Starting Price | 🎯 Best For | ⚡ Key Pricing Advantage | |---------|------------------|------------|--------------------------| | [LiveKit](https://livekit.com) | $0/month (Build); [$50/month](https://livekit.com/pricing) (Ship) | End-to-end voice AI covering agent deployment, inference, telephony, and WebRTC transport under one bill | One billing account covers the full stack | | [VAPI](https://vapi.ai) | [$0/month](https://vapi.ai/pricing) (usage-based); Enterprise requires annual contract | Developer teams that want a managed voice agent layer with BYOK for STT/TTS/LLM | $0.05/min platform fee; $0 model costs with your own API keys. HIPAA is a $2,000/month add-on | | [Pipecat Cloud](https://pipecat.daily.co/sign-in) | [$0.01/min](https://www.daily.co/pricing/pipecat-cloud/) (agent-1x, voice); $0.02/min (agent-2x, voice + video) | Developer teams building on the open-source Pipecat framework need managed hosting | Infrastructure billed per minute; model costs passed through at cost with BYOK option | | [Retell AI](https://www.retellai.com) | [$0/month](https://www.retellai.com/pricing) ($10 free credits to start); then $0.07+/min | Voice AI teams that want a fully managed platform and pay only for minutes used | No platform fee, 20 concurrent calls included, pay only for what you use |

LiveKit and VAPI charge a monthly base fee on top of usage. Pipecat Cloud and Retell AI bill purely by the minute with no monthly floor. Compare total cost at your expected volume, not just the entry price.

## LiveKit vs Retell AI: Which Should You Choose? LiveKit gives you control over each layer of the stack, while Retell AI pre-wires the pipeline and charges a **single per-minute rate** that bundles the core components. Which one fits depends on how much of that control you need. **LiveKit is better for:** - Building multimodal agents that need voice, video, and WebRTC transport as a core requirement - Granular control over STT, TTS, LLM, and media transport without being locked into a pre-wired pipeline - Organizations on Scale that need HIPAA, region pinning, custom voice cloning, and inference discounts in a single plan **Retell AI is better for:** - Going from signup to a live voice agent in minutes without wiring up separate vendors - Projects where cost predictability matters more than control, since Retell charges only for minutes used - Voice-only deployments where SOC 2 compliance is needed from day one, and the plan costs less than $500/month **Use both if** you're running LiveKit for WebRTC-based applications and want Retell to manage a dedicated telephony calling pipeline independently. **The two serve different functions**, so running them together doesn't create overlap. ## LiveKit and Cekura: Two Tools, One Stack LiveKit covers agent deployment, inference, telephony, and WebRTC transport. Automatic scenario testing, ASR error detection, and **production monitoring on the voice layer** sit outside what it does. That's where [Cekura](https://www.cekura.ai) comes in. Many teams reach launch with the infrastructure running cleanly, but **the voice layer is untested** under production load. Running both gives you a better shot at catching those failures before they reach users. **Use LiveKit when you need:** - Agent deployment, inference, telephony, and WebRTC transport in a single managed platform - Session-level observability with transcripts, recordings, and latency metrics per conversation **Use Cekura when you need:** - Pre-production testing - Pre-production simulation of your voice and chat agent across hundreds of scenarios - A/B testing to [compare](https://www.cekura.ai/blogs/cekura-a-b-testing-platform-dialog-strategies-in-voice) multiple versions of your agent against the same call scenarios in one place - Multi-turn red teaming to stress-test against jailbreaks, bias, and adversarial attacks - Custom personas and accents, where you can test against Cekura's curated caller library or build your own with specific accents and background noise - Production monitoring - Production monitoring for latency, instruction-following, tool-call accuracy, and [CSAT](https://www.cekura.ai/blogs/cekura-a-complete-conversational-ai-testing-suite) on live calls - Slack alerts for latency spikes and quality drops, so you find out before your callers do - Pipeline & compliance - [QA runs](https://docs.cekura.ai/documentation/guides/testing-agents/overview) in your CI/CD pipeline, triggered on prompt changes - Cron job scheduling to run tests on a recurring schedule and catch silent regressions between deployments - Preset and custom metrics with native latency, instruction-following, and tool-call tracking, plus fully configurable LLM judges - Native integrations with [Retell](https://docs.cekura.ai/documentation/integrations/retell/testing), [VAPI](https://docs.cekura.ai/documentation/integrations/vapi/testing), [ElevenLabs](https://docs.cekura.ai/documentation/integrations/elevenlabs/testing), [Pipecat](https://docs.cekura.ai/documentation/integrations/pipecat/automated), [Bland](https://docs.cekura.ai/documentation/integrations/chat-testing#bland), and more - **[SOC 2](https://www.linkedin.com/posts/sidhantkabra_cekura-yc-f24-is-now-soc-2-type-ii-compliant-activity-7331019517831098368-OX7N)-, HIPAA-, and GDPR-compliant:** Transcript redaction, role-based access, and audit trails. **Use both if** you're shipping a voice AI product to production. LiveKit covers the infrastructure layer. Cekura covers how your voice layer holds up under production conditions. Together, the **two tools cover what neither handles** on its own. If you're already running LiveKit and need to cover the voice layer too, Cekura connects to your existing stack in under 15 minutes. [Request a demo](https://www.cekura.ai/expert) to see how it fits. ## My Bottom Line on LiveKit Pricing Build is worth trying before you pay anything. Ship at $50/month works once you're taking a **voice agent to production**, as long as you don't have compliance requirements and your traffic is predictable enough to model the overage costs. Scale at $500/month makes sense when **your monthly volume justifies the floor**, or the moment HIPAA and SOC 2 become non-negotiable. Enterprise is for teams that've outgrown Scale's published limits and need a formal agreement to match. ## Frequently Asked Questions ### How Much Does LiveKit Cost? LiveKit pricing **starts at $0/month on the Build plan**. Paid plans start at $50/month (Ship) and $500/month (Scale), with Enterprise at custom pricing. All plans are usage-based, so your **actual bill depends on agent session minutes**, inference usage, and telephony volume. ### What Is the Difference Between LiveKit Build, Ship, and Scale? The main difference between Build, Ship, and Scale is what **happens once you've hit a limit**. Build operates on hard caps with no overages. Ship converts those caps into soft limits billed per unit. Scale adds HIPAA, SOC 2, inference discounts, and up to **600 concurrent agent sessions** for teams with sustained production traffic. ### Does LiveKit Have a Free Plan? Yes, **LiveKit has a free plan called Build**. It includes 1,000 agent session minutes, $2.50 in inference credits, and one **free US local phone number per month**, with no credit card required. ### What Are LiveKit Agent Session Minutes? Agent session minutes **measure the active time your deployed agent is connected to a session**. Billing starts when the agent joins the room and **stops when the agent disconnects** or the room ends. The Build plan includes 1,000 minutes per month; Ship includes 5,000; Scale includes 50,000. ### Is LiveKit Open Source? Yes, LiveKit's **media server and agent framework are fully open source** and available to self-host. LiveKit Cloud **is the managed version**, with managed agent deployment, observability, and global infrastructure. ### Does LiveKit Pricing Include HIPAA Compliance? **HIPAA compliance is included** in the Scale plan at $500/month and on Enterprise. Scale is the **minimum tier for regulated industries**, with Enterprise available for teams that need custom terms on top.

Posts tagged with "Ai Infrastructure"

LiveKit Pricing Explained: Plans, True Costs, and Which Tier Fits