LiveKit Cloud pricing starts at $0/month for Build, $50/month for Ship, and $500/month for Scale. Enterprise pricing is custom. But the plan fee is only part of the cost: LiveKit also meters agent sessions, WebRTC usage, inference, telephony, data transfer, recording, and other resources.
For AI voice agents in particular, your final cost depends heavily on the models you use for LLM, speech-to-text (STT), and text-to-speech (TTS).
Here’s how LiveKit pricing works in 2026 and what to budget for.
LiveKit Pricing at a Glance
| Plan | Monthly price | Agent session allowance | Inference credits | Main use |
|---|---|---|---|---|
| Build | $0 | 1,000 min | $2.50 | Development and testing |
| Ship | $50 | 5,000 min | $5 | Production applications |
| Scale | $500 | 50,000 min | $50 | Larger deployments |
| Enterprise | Custom | Custom | Custom | High-volume/enterprise |
Build requires no credit card. Ship and Scale include larger resource allowances and additional platform features, while Enterprise adds custom pricing and support terms.
What Does LiveKit Actually Charge For?
LiveKit uses resource-based billing, rather than charging one universal per-minute rate.
The main cost categories are:
- Agent session minutes — time an agent is actively connected to a session.
- WebRTC usage — realtime media and participant connections.
- AI inference — LLM, STT, and TTS usage.
- Telephony — phone numbers and SIP/call usage.
- Data transfer — media and other data transferred through the platform.
- Recording and export — recorded or exported media.
- Observability — agent transcripts, logs, observations, and related resources.
LiveKit’s documentation says each metered resource has its own allowance and billing unit. On paid plans, usage beyond the included allowance is billed at the published rate; Build uses hard limits rather than paid overages.
That means the same $50 Ship plan can produce very different total bills depending on the application.
LiveKit Plan Details
Build: $0/month
Build is LiveKit’s free tier for starting a project. It includes agent deployment, observability, inference credits, a global edge network, one free phone number, session metrics and analytics, and community support. The plan includes 1,000 agent session minutes and $2.50 in inference credits.
Build is best suited to development, testing, and smaller workloads. Once a metered allowance is exceeded, the Build plan uses a hard cap rather than automatically billing additional usage.
Ship: $50/month
Ship is the first paid plan and is designed for applications being shipped to real users. It adds team collaboration, custom voices, deployment rollback, and email support. It also increases the included usage allowances compared with Build.
Ship includes 5,000 agent session minutes and $5 in inference credits. Additional usage is billed according to the applicable resource rates.
Scale: $500/month
Scale is aimed at larger applications and teams that need additional access, security, and regional controls. It adds role-based access, metrics export APIs, region pinning, security reports/HIPAA eligibility, and inference discounts. It includes 50,000 agent session minutes and $50 in inference credits.
If your deployment requires LiveKit’s published HIPAA eligibility or region-pinning capabilities, Scale is the relevant published tier to evaluate.
Enterprise — Custom
Enterprise includes the Scale feature set plus options such as volume pricing, SSO, a shared Slack channel, and a support SLA. Pricing is negotiated based on requirements and usage.
How Much Does a LiveKit Voice Agent Cost?
There isn’t one fixed LiveKit voice-agent price.
A typical voice-agent calculation can include:
Agent session + LLM + STT + TTS + telephony + observability
LiveKit’s current pricing calculator displays an example configuration using:
- Agent session: $0.01/min
- Telephony: $0.01/min
- LLM: $0.0014/min
- STT: $0.0058/min
- TTS: $0.03/min
- Observability: $0.01/min
That example adds up to roughly $0.067 per active minute, before considering the monthly plan fee and any other applicable usage. It should be treated as a calculator example, not a universal LiveKit voice-agent rate.
The reason is simple: LiveKit Inference prices different model types differently. LLMs are billed by tokens, STT by duration, and TTS by characters. Model selection therefore changes the final cost.
Example: 10,000 minutes of voice-agent usage
Suppose an application handles 10,000 minutes of agent sessions in a month. You cannot simply multiply 10,000 by one universal LiveKit rate.
You need to account for:
| Cost | What affects it |
|---|---|
| Plan | Build, Ship, Scale or Enterprise |
| Agent sessions | Total connected agent time |
| LLM | Tokens and selected model |
| STT | Audio duration and selected model |
| TTS | Characters and selected model |
| Telephony | Calls and phone numbers |
| Observability | Events and recordings |
| Data transfer | Media/data volume |
This is why LiveKit’s own calculator is more useful than a single advertised “per-minute” number.
LiveKit Inference Pricing
LiveKit Inference gives developers access to LLM, STT, and TTS models through a unified interface. The included inference credits are:
- Build: $2.50
- Ship: $5
- Scale: $50
LiveKit currently estimates these credits at roughly 50, 100, and 1,000 minutes respectively, but the actual duration depends on the models being used.
Scale also provides discounted rates for many STT and TTS models, while Enterprise customers can negotiate preferred high-volume rates.
This model-based pricing is important if you’re comparing LiveKit with platforms that bundle AI costs into a single per-minute rate.
LiveKit vs Agora vs Twilio
LiveKit, Agora, and Twilio don’t price exactly the same product, so headline rates should not be treated as direct equivalents.
| Platform | Starting price | Pricing approach | What you’re primarily paying for |
|---|---|---|---|
| LiveKit Cloud | $0 / $50 / $500 | Plan + usage | Realtime infrastructure, agents and optional inference |
| Agora Conversational AI | $0.10/min | Usage-based | Conversational AI audio tasks |
| Twilio Video | $0.004/participant-min | Pay-as-you-go | Video room participant time |
| Self-hosted LiveKit | No Cloud plan fee | Infrastructure | Your own hosting and operations |
Agora’s Conversational AI Platform currently starts at $0.10/min, with the first 300 minutes free. Its published price includes selected ASR, LLM, and TTS usage, making it a simpler unit-price comparison for conversational AI.
Twilio Video starts at $0.004 per participant-minute with no subscription fee. However, that is a Video product rate—not an equivalent all-in-one AI-agent rate. Recordings, compositions, transcription, and voice connectivity have additional pricing.
So the right comparison depends on what you’re building:
- AI voice/realtime agents: Compare the complete agent stack.
- Plain realtime video: Participant-minute pricing becomes more relevant.
- Large-scale AI: Model costs and volume discounts deserve more attention.
- Self-hosted infrastructure: Compare Cloud fees against infrastructure and engineering costs.
LiveKit Cloud vs Self-Hosting
LiveKit’s media server and Agents framework are open source, so teams can run the technology on their own infrastructure. That removes LiveKit Cloud’s managed-service fees, but it doesn’t make the system cost-free.
With self-hosting, you take responsibility for:
- Compute
- Networking
- Storage
- Scaling
- Monitoring
- Security
- Deployment
- Reliability
- Engineering operations
For predictable, high-volume workloads, self-hosting can be worth evaluating. For smaller teams, the engineering and operational overhead may be more important than the infrastructure invoice itself.
The key comparison is therefore:
LiveKit Cloud cost vs infrastructure + engineering + operations.
Which LiveKit Plan Fits Your Use Case?
| If you need… | Plan to evaluate |
|---|---|
| Development and testing | Build |
| A small production deployment | Ship |
| Higher scale or advanced access/security features | Scale |
| HIPAA eligibility, region pinning or RBAC | Scale |
| Custom enterprise pricing and support | Enterprise |
The plan should be selected based on both usage and required features. A team with low traffic but a requirement for certain security features may need a higher tier even if its raw usage would fit a cheaper plan.
LiveKit Pricing: What to Budget For
Before estimating your monthly cost, check these six areas:
- Plan fee — $0, $50, $500, or custom.
- Agent session usage — how many minutes agents stay connected.
- AI models — LLM, STT, and TTS selection.
- Telephony — phone numbers and call usage.
- Media/data usage — WebRTC, transfer, recording, and export.
- Observability — transcripts, logs, recordings, and other monitored resources.
The biggest mistake is estimating LiveKit purely from the monthly subscription price.
Bottom Line
LiveKit Cloud starts at $0/month, but your actual cost depends on usage. For AI voice applications, the important variables are agent session time, model selection, telephony, WebRTC/data usage, and observability not just the monthly plan.
For a prototype, Build provides a way to start without a subscription. Ship is the first paid tier for production use, while Scale becomes relevant when higher usage or advanced access, security, and regional features are required.
Before choosing a plan, model your expected minutes + AI models + telephony + other metered resources. That will give you a much more realistic LiveKit cost estimate than looking at the plan price alone.
Is LiveKit free?
Yes. LiveKit Cloud has a $0 Build plan with included usage allowances. Build has hard caps on its included resources rather than paid overages.
How much does LiveKit cost?
LiveKit Cloud currently starts at $0/month for Build, $50/month for Ship, and $500/month for Scale. Enterprise pricing is custom. Additional usage can increase the final bill.
Does LiveKit charge separately for AI inference?
Yes. LLM, STT, and TTS usage are metered according to their respective pricing units and models.
How much does a LiveKit voice agent cost per minute?
There is no universal rate. LiveKit’s current calculator shows an example configuration of about $0.067/min, but the actual cost depends on the selected AI models, telephony, agent usage, observability, and other resources.
Is LiveKit cheaper than Twilio?
It depends on the use case. Twilio Video’s $0.004 participant-minute price is for Video and should not be compared directly with the total cost of a LiveKit AI voice agent.
Is LiveKit cheaper than Agora?
The pricing models are different. Agora Conversational AI currently lists $0.10/min for its Conversational AI audio task, while LiveKit separates platform usage and inference costs. Your workload and model selection determine the actual comparison.
Can I self-host LiveKit?
Yes. Self-hosting lets you run LiveKit’s open-source components on your own infrastructure, but you take responsibility for hosting, scaling, security, and operations.
Pricing checked in September 2026. LiveKit, Agora, and Twilio can change pricing, quotas, and model availability. Verify current pricing directly with the providers before making a purchasing decision.
Ajay Kumar is an SEO and AI search specialist with more than 10 years of experience across SEO, app store optimisation and generative engine optimisation. He has led search growth for gaming, software, fintech and healthcare brands in India and internationally, and he edits every CompareCrest comparison, checking pricing and claims against vendor sources before publishing.
