We may earn a commission if you buy through links on this page. It never changes a score or a ranking. How we make money
The bottom line
A strong free starting point for developer-led teams building real-time AI agents.
A small team building a voice or video AI agent can start on LiveKit’s free Build plan, which includes 1,000 agent session minutes and five concurrent sessions. The open-source framework and cloud platform cover agent development, deployment, telephony, and session observability.
The trade-off is that LiveKit is a developer platform, not a no-code agent builder. Ship costs $50 per month, with usage-based charges beyond some included limits. Teams without engineering time may find setup and ongoing model choices harder than the price suggests.
Good fit if
- Developers building real-time AI agents
- Teams testing voice agent ideas
- Projects needing telephony and SIP
Skip it if
- Teams seeking a no-code agent builder
- Hobby projects with minimal budgets
- Teams without developer capacity

ScoresHow it rates
What works
- Free Build plan includes 1,000 agent minutes
- Open-source framework offers agent customization
- Cloud deployment and scaling are available
- Telephony, SIP, and phone numbers are supported
- Session observability helps inspect agent activity
What doesn't
- Setup can require command-line and server knowledge
- Ship's included minutes are capped at 5,000
- Model usage beyond credits costs extra
- Several telephony and recording charges are usage-based
- Annual billing rates are not listed
01Who LiveKit is for
LiveKit is for small, developer-led teams building voice, video, or physical AI agents—not for a team that just needs a basic meeting app. Its open-source framework lets developers build and customize agent behavior and model choices. LiveKit Cloud handles deployment and scaling, with telephony, SIP integrations, and session observability among the listed features.
For a team of one to ten, the free Build plan makes it possible to start without a monthly platform fee. That can suit an engineer testing an idea or a small product team validating a real-time agent. Once the team is serving users, Ship is $50 per month before usage charges. The budget question is less about seats—LiveKit lists flat plan prices, not per-seat rates—and more about session minutes, model use, and the engineering time needed to build and maintain the agent.
02LiveKit free plan and trial
LiveKit’s free Build plan includes 1,000 agent session minutes per month, five concurrent agent sessions, one agent deployment, $2.50 in inference credits, and one free US local number. The plan is a meaningful starting point for prototyping or a small agent with modest use. LiveKit does not publish a free trial separately in the published pricing details; Build is the free entry point.
Build’s listed inclusions are limited, and the pricing page does not list paid overage rates for Build. For a live service that needs more included capacity, Ship costs $50 per month and includes 5,000 session minutes, 20 concurrent sessions, two deployments, $5 in inference credits, and 5,000 recording minutes. Ship session minutes beyond the included amount cost $0.01 per minute. Check expected model and telephony use before moving a customer-facing agent onto a paid plan.
03What a small team pays for LiveKit
LiveKit lists flat monthly plan prices, not per-seat charges. Build costs $0 per month whether one person or five people are working on the project. Ship is $50 per month: that works out to $50 for one person, $16.67 per person for a three-person team, or $10 per person for five, as simple budget comparisons—not seat pricing. The pricing page does not show annual billing or annual rates.
Ship includes 5,000 agent session minutes, then charges $0.01 per minute; it also includes 5,000 recording minutes, then charges $0.005 per minute. Other potential extras include inference beyond the $5 credit at model prices, US local numbers at $1 per month after the included number, and US local inbound minutes at $0.01 per minute over included amounts. Audio turn detection is $0.50 per 1,000 requests where charged. A three- or five-person team should estimate minutes and model use before treating the $50 base price as its full bill.
04Getting started with LiveKit
LiveKit’s framework gives developers control over agent behavior and model choices, while its cloud platform provides deployment and scaling. The listed toolkit also includes a model gateway, telephony integrations, automatic turn detection, interruption handling, and session observability. That is a broad starting set for a voice agent, but it still calls for development work to connect an agent to a product and choose its models.
Reddit reports on setup vary. One commenter said getting a server running took command-line knowledge; another described the instructions as lengthy. A LiveKit Cloud user in a Foundry setup said it was quick and easy, while a developer working on an existing app said they struggled to find clear integration documentation. These are individual experiences, not a setup guarantee. For day one, plan for an engineer to connect the client, configure the agent, and decide whether to use Cloud or self-host the open-source framework.
05When small teams outgrow LiveKit
A team may need to move beyond Build when its usage exceeds 1,000 session minutes a month, it needs more than five concurrent sessions, or it needs additional deployment capacity. Ship raises those included limits to 5,000 minutes, 20 concurrent sessions, and two deployments for $50 per month. Scale is $500 monthly and includes 50,000 minutes and up to 600 concurrent sessions. Ship and Scale charge $0.01 per session minute beyond their included amounts.
The next pressure point is the bill’s variable portion: model usage, recordings, phone numbers, inbound calls, and other listed services can add charges. Teams may also outgrow the developer-led approach if they lack engineering time for integrations and agent changes. Reddit users describe both successful use and setup friction: some report easy Cloud configuration, while others say self-hosting takes command-line knowledge and a capable server. Ask LiveKit about expected usage costs and support needs before scaling a production workload.
From the threadsWhat users say
Reddit sentiment is mostly positive from users who got LiveKit working, with praise for call quality and the free tier. Setup and integration can be frustrating, especially for self-hosters or developers connecting an existing app.
easy to setup, and the free tier has enough for me running 4 sessions a month with 5 players plus myself for about three hours per Session. Works like a charm
Livekit works great, but you've got to have some command line savvy to get it working, along with a central server on a good connection
It was quick and easy to make the changes with livekit.io setup providing everything I needed to get the custom service set up and running
when I tried integrating It with the application, I was unable to find any proper document to do it
Questions people ask
Is LiveKit free?
Yes. The pricing page lists Build at $0 per month. It includes 1,000 agent session minutes, five concurrent sessions, one deployment, $2.50 in inference credits, and one free US local number. Build’s listed limits make it most suitable for testing or low-volume use.
How does LiveKit work?
LiveKit combines an open-source framework for building and customizing agents with a cloud platform for deploying and scaling them. Its listed features include real-time voice, video, telephony, a model gateway, session observability, and automatic turn and interruption handling.
Can we self-host LiveKit?
The LiveKit framework is open source, and Reddit users describe running their own LiveKit servers. Self-hosting means supplying and managing a server; commenters mention command-line knowledge and a good connection as considerations. LiveKit Cloud is the hosted option.
How much does LiveKit inference cost?
Build includes $2.50 in inference credits, Ship includes $5, and Scale includes $50. The pricing details say usage beyond those amounts is billed at model prices, which vary by model and plan; Scale includes discounted model prices.
Does ChatGPT use LiveKit?
A Reddit poster cited a claim on LiveKit’s site that OpenAI uses LiveKit to deliver voice to users. That statement is not independently confirmed in the information available here, so teams should ask LiveKit for current details if this affects a buying decision.
Sources
- LiveKit homepage
- LiveKit pricing page
- Reddit: Anyone have experience with LiveKit for voice and if it's better than the webRTC?
- Reddit: Livekit : realtime multimodal conversation with AI.
- Reddit: Lost between LiveKit Cloud vs Vapi vs Retell for a voice AI agent (~3,000 min/month) – real costs & recommendations in 2025?
- Reddit: Introducing LiveKit - a WebRTC video conferencing server in Go
