LiveKit pricing: Cloud plans, agent costs and reviews
LiveKit Cloud shows $0.0100 per agent-session minute in its Build and Ship estimate. Its calculator lists telephony and model inference separately.

On this page
LiveKit pricing and what it covers
LiveKit Cloud shows $0.0100 per agent-session minute in its Build and Ship estimate. Its calculator lists telephony and model inference separately. LiveKit vendor source.
Figures in the original research below are dated examples. Use the vendor’s current terms and the same workload when estimating your cost.
LiveKit Cloud meters usage across bundled plans. This is what each plan commits to and includes, alongside the self host option.
1,000 agent minutes, 5,000 WebRTC minutes, 50 GB of data and $2.50 of inference credit, with 5 concurrent sessions and no credit card.
5,000 agent minutes, 150,000 WebRTC minutes, 250 GB and $5 of inference, with 20 concurrent and $0.01 a minute agent overage.
50,000 agent minutes, 1.5M WebRTC minutes, 3 TB and $50 of inference, with up to 600 concurrent and discounted model rates.
Run the open source server on your own cloud, bring your own models, and pay your compute, bandwidth and operations.
Custom limits, private links, dedicated support and a business associate agreement, priced by contract.
Rates checked on LiveKit's pricing and billing pages on 21 October 2026. Plans are committed buckets, so once you pass Build you pay the plan base whether or not you use every bundled minute.
Estimate your total cost
For voice-agent infrastructure, add fixed fees, metered usage and any separately priced services. Use the allowance in your plan and price any usage above it.
| Input | What to record |
|---|---|
| Workload | voice-agent infrastructure |
| Billable items | Managed session time, transport, model inference and the cost of running your deployment. |
| Terms | Currency, billing period, included usage, overages and contract minimum. |
| Pilot | Completion rate, connected time and operating effort. |
Figures in the original research below are dated examples. Use the vendor’s current terms and the same workload when estimating your cost.
A LiveKit voice agent has three parts: the plan base, the per minute platform fee, and the model cost. The model cost is separate from the agent minute.
What to test with LiveKit
For speech inside your existing framework, follow the LiveKit integration guide and runnable starter or its Markdown guide for agents. The Python adapter is maintained by PyAI; this does not imply inclusion in the upstream provider directory. Read current PyAI pricing and verify your project identity and usage limits before running a sample.
Measure the same session with your chosen carrier, inference provider and deployment, including connected time.
Build one end-to-end session, then test reconnection, interruption and logging under the expected load.
The free tier covers up to 1,000 agent minutes a month, enough to prototype a voice agent.
$50 a month with 5,000 agent minutes, then $0.01 a minute, for a launched app.
$500 a month with 50,000 agent minutes and higher concurrency, for steady production.
No platform fee, but you run Kubernetes, operations and your own models, best past heavy scale.
Private links, custom limits and a business associate agreement, negotiated by contract for regulated workloads.
LiveKit pricing calculator: estimate your monthly agent bill
LiveKit figure is the plan base plus $0.01 a minute platform over the bundle plus about $0.05 a minute inference, from its own credit rate. It excludes telephony and varies by model. PyAI figure is minutes times the all in per minute rate, with no plan base.
Fees and terms to confirm
- Ask which charges apply for managed session time, transport, model inference and the cost of running your deployment.
- Confirm what is included in the base plan and what is metered separately.
- Ask how the vendor calculates minimums, overages and renewals.
- Models are extra: the $0.01 a minute agent fee is the platform only. Speech to text, the model and text to speech bill on top, about $0.05 a minute through inference credits.
- 10 second minimums: WebRTC, SIP, agent and recording sessions each carry a 10 second minimum, which inflates short calls.
- Per resource round up: each resource rounds up to its minimum increment before the bill is summed, so small usage is rounded generously.
- Downstream data: bandwidth is billed on downstream data at $0.10 to $0.12 a gigabyte on paid plans, not upstream.
- Metering until shutdown: agent metering continues until the room ends or the agent disconnects, so a missed shutdown call keeps billing.
- Committed buckets: once past Build you pay the $50 or $500 plan base whether or not you use every bundled minute.
- Telephony rentals: phone numbers carry a monthly rental on top of per minute call charges.
Read customer evidence for your workload
Look for recent customer examples using the same product edition and deployment. Ask the vendor for a reference or a pilot if the published example leaves out your workflow.
Only 1 review on G2
No meaningful Capterra or Trustpilot base
What developers like
What developers dislike
LiveKit has a thin formal review base: one G2 review at 4.5, and no meaningful Capterra or Trustpilot listing, checked October 2026. The sentiment below is from labeled developer and vendor blogs, not neutral reviews, so weigh it accordingly.
What reviewers say about LiveKit pricing
Strong on scalability and features, though the pricing was a concern worth watching as usage grows.
The low per minute cost is real, but the bigger cost is the long project to get a voice agent to production.
Comments paraphrased from a public G2 review and a labeled vendor blog, linked to the source pages. Formal user reviews for LiveKit are sparse.
What developers love
Developers value the open source flexibility and low latency of the stack.
Where developers struggle
There is a steeper learning curve and a thin presence on traditional review sites.
Developers value the open source flexibility, low latency and clean metering, but note a steep learning curve, a long build to production, and committed plan buckets.
Compare the same scope
The managed-agent comparison below concerns PyAI Omni. For PyAI speech inside LiveKit, keep your existing stack and budget Hear and Speak independently alongside LiveKit, your LLM and any telephony.
Keep the workload, channel and required support consistent across options. A price for one component covers a different scope from a proposal for a complete deployment.
For PyAI’s billed products and telephony terms, see current PyAI pricing.
| Criteria | LiveKit | PyAI |
|---|---|---|
| What it is | Infrastructure plus a framework you assemble | ✓ A hosted voice API, models included |
| Pricing model | Plan base plus metered usage | ✓ Per minute, no plan base |
| Agent cost all in | About $0.06/min plus plan base | See current PyAI pricing |
| Models | Bring your own or buy Inference | ✓ Hosted and included in the rate |
| WebRTC transport | ✓ Global edge media network | No WebRTC transport |
| Self hosting | ✓ Open source, self host the server | Hosted API only |
| Setup effort | Assemble and operate the stack | ✓ Call one API, OpenAI compatible |
| Phone ready audio | Supported | ✓ PyAI delivers WAV, PCM or G.711 straight from the API |
| Speech to text and TTS | Via your models or Inference | See current PyAI pricing |
| Reviews | Thin, 1 G2 review | No G2 profile yet |
Choose LiveKit if you want
- WebRTC transport over a global edge media network for low-latency live audio in browsers and mobile apps.
- An open source media server you can self-host on your own cloud account, paying only for compute.
- Full control to bring your own STT, LLM and TTS and wire them as you like, including local or fine-tuned models.
Choose PyAI if you want
- See current PyAI pricing
- Per-minute pricing metered by the second, with no plan base to build a budget around.
- Phone calls handled natively at 8 kHz, with managed numbers in the US and India.
Check normal and peak usage
Estimate one normal billing period and one peak period. In each scenario record your billable units, included allowances, overages and any fixed fees. Compare the totals per completed task.
Figures in the original research below are dated examples. Use the vendor’s current terms and the same workload when estimating your cost.
A LiveKit voice agent bill depends on the plan base plus usage. These estimates use the platform fee plus about $0.05 a minute inference, and exclude telephony.
| Use case | What you need | Plan to pick | Watch out for |
|---|---|---|---|
| Hobby and prototype | Under 1,000 agent minutes | Build ($0) | Inference beyond the small credit |
| Small production | About 5,000 agent minutes | Ship ($50) | Models and telephony on top |
| High volume | 50,000 agent minutes or more | Scale ($500) | Overage and committed base |
LiveKit vs PyAI monthly agent cost by volume
LiveKit is the plan base plus $0.01 platform plus about $0.05 inference a minute. PyAI Omni is minutes times the all in rate, with no plan base. The two land close per minute, so weigh control against a hosted stack.
| Agent minutes a month | LiveKit plan | LiveKit all in | PyAI Omni |
|---|---|---|---|
| 1,000 | Build | ~$60 | See current PyAI pricing |
| 5,000 | Ship | ~$300 | See current PyAI pricing |
| 50,000 | Scale | ~$3,000 | See current PyAI pricing |
| 100,000 | Scale plus overage | ~$6,000 | See current PyAI pricing |
LiveKit figures use the platform fee plus about $0.05 a minute inference from its own credit rate, before telephony, and vary by model. At these volumes the two are close per minute, and LiveKit can run a little lower, so the choice is control and self hosting versus a hosted stack. See current PyAI pricing
What to take into a vendor conversation
Measure the same session with your chosen carrier, inference provider and deployment, including connected time.
LiveKit pricing runs from a free Build tier to Ship at $50 and Scale at $500 a month, plus custom Enterprise, with agent sessions at $0.01 a minute over the bundle and model costs on top. All in, a Cloud voice agent is near $0.06 a minute plus the plan base, and self hosting removes the platform fee for your own operations. Check your selected project identity, billing status and effective limits before testing. See current PyAI pricing
Keep LiveKit for WebRTC transport, its open source agent framework and worker hosting. Add PyAI Hear and Speak through the Python adapter when you want PyAI speech inside that application. Choose Omni separately when you want a managed conversation runtime.
LiveKit pricing FAQs
How much does LiveKit cost?
LiveKit Cloud shows $0.0100 per agent-session minute in its Build and Ship estimate. Its calculator lists telephony and model inference separately.
What should I include in a LiveKit estimate?
Include managed session time, transport, model inference and the cost of running your deployment. Use the same billing period across all quotes and ask how excess usage is charged.
How can I test LiveKit before committing?
Measure the same session with your chosen carrier, inference provider and deployment, including connected time. Build one end-to-end session, then test reconnection, interruption and logging under the expected load.
Is LiveKit free?
LiveKit is open source, so you can self host the server and pay only your own cloud infrastructure, with no LiveKit platform fee. LiveKit Cloud also has a free Build tier at $0 a month with 1,000 agent session minutes and no credit card. Beyond the Build limits you move to Ship at $50 or Scale at $500 a month.
Confirm the current rate and what it includes with the vendor.
How are LiveKit agent minutes billed versus WebRTC minutes?
LiveKit meters agent sessions and WebRTC media separately. Cloud deployed agents bill as agent session minutes, at $0.01 a minute over the bundle. Self hosted agents instead count against WebRTC participant minutes. Model costs for speech to text, text to speech and the language model are billed on top, through LiveKit Inference credits or your own provider keys.
Confirm the current rate and what it includes with the vendor.
What does a LiveKit voice agent really cost per minute?
On LiveKit Cloud, a voice agent is about $0.01 a minute for the platform plus model costs. LiveKit Inference credits imply about $0.05 a minute for speech to text, the model and text to speech, so an all in rate near $0.06 plus the plan base and telephony.
Confirm the current rate and what it includes with the vendor.
What are LiveKit's hidden costs?
LiveKit rounds each resource up to its minimum increment before adding them, and applies a 10 second minimum per session, which inflates short calls. Data is billed on downstream bandwidth at $0.10 to $0.12 a gigabyte. Agent metering continues until the room ends or the agent disconnects, so a missed shutdown keeps billing.
Confirm the current rate and what it includes with the vendor.
What is a hosted alternative to assembling LiveKit?
PyAI Omni provides a managed conversation runtime. You can also keep LiveKit and use PyAI Hear and Speak through livekit-plugins-pyai. The adapter preserves your rooms, worker and provider choices; LiveKit, hosting, your LLM and any telephony remain separate from PyAI speech usage.
Can I use PyAI inside LiveKit Agents?
Yes. The PyAI-maintained livekit-plugins-pyai Python package supplies Hear STT and Speak TTS inside AgentSession. Keep your LiveKit room, worker, LLM, tools and turn policy; no Omni agent is required. PyAI speech usage is billed separately from LiveKit, worker hosting, your LLM and any telephony.
Related pricing and comparisons
Estimate PyAI phone costs
This calculator estimates PyAI charges for the phone setup you select. Use the pricing sections above for other providers. See the PyAI rate card for billing terms.
- AI usage · 150 min
- $7.50
- Managed telephony · 200 billable min
- $2.00
- Estimated monthly total
- $9.50
AI usage bills by the second. Managed telephony bills in whole-minute increments per call. This estimate assumes every call lasts the duration you enter.