Introducing the LiveKit Startup Program: Your whole stack, on us

LiveKit lets you build an agent from scratch in about ten minutes. Simply run lk agent init, and the starter template gives you a working model stack with a single LiveKit API key.

But a working stack isn't necessarily ready to be customer-facing. Speech, language, and voice models all affect one another, and so does everything around them: turn detection, keyterm prompting, noise cancellation, and voice selection. None of this is fully captured in benchmarks; the best voice agents need to be tuned based on real conversations.

Experimentation, however, comes with overhead. Doing things right means that pretty soon you have multiple API keys in your .env, separate billing systems with different rates and limits, and several places to look when something breaks mid-demo, all of which slow down mature teams, let alone startups looking to get their first agents to market.

We built the LiveKit Startup Program to help early stage companies solve this problem.

The LiveKit Startup Program gives eligible startups $7,000 in inference credits across seven partners, $10,000 in LiveKit Cloud credits, and one year on the Scale plan (normally $500 per month) to ensure that early stage companies have the resources they need to build highly performant agents that feel enterprise-grade. All told, it adds up to about $23,000 in total value devoted to the most critical period of agent refinement.

One application gives access to the entire program. Models from participating partners are available through LiveKit Inference, with no separate provider accounts, API keys, or bills to manage. You can experiment across the stack while keeping the same integration, then take what works into production on LiveKit Cloud.

What you get#

We're onboarding more inference partners across the real-time AI stack, and we'll keep adding credits to the program as they join it. Each pool of credits draws down on its own, and if one runs out, that service simply rolls over to standard rates so nothing in your agent stops working.

All inference credits are redeemable through LiveKit Inference with no separate accounts to create, applications to fill out, or provider keys to manage. Your models are strings in your agent config, and the corresponding program credits are applied automatically.

1
session = AgentSession(
2
    stt="assemblyai/universal-3-5-pro:en",
3
    llm="google/gemma-4-31b-it",
4
    tts="inworld/inworld-tts-2:Ashley",
5
)

Try a different model by changing one string, without changing the rest of your agent.

One year of LiveKit Cloud Scale included#

Once you've used your inference credits to find the right models, the program also covers the next step: putting them into production on LiveKit Cloud Scale (normally $500 per month). Beyond the plan's included usage, you also receive $10,000 in LiveKit Cloud credits, giving you room to grow before paying anything out of pocket.

Scale includes 50,000 agent session minutes per month, up to 600 concurrent agent sessions, region pinning, role-based access control, metrics export APIs, custom voices, and discounted inference rates for any provider not participating in the program.

Better yet, you don't have to rebuild the operational or security layer as you move from experimenting to production.

Everything runs on security-first, SOC 2 Type II compliant infrastructure. Zero data retention (ZDR) is enabled by default for every model in LiveKit Inference with no configuration required. Your prompts, audio, and model outputs are processed only to generate a response. They are never stored or used for training (you can read more about how LiveKit enforces zero data retention).

This means when a prospective customer asks how their data is protected, you have one clear answer across every model in LiveKit Inference without chasing down each provider.

Get started building#

The program is for startups with fewer than 50 employees that have raised less than $5 million and are making voice, video, or physical AI a core part of their product. You should be incorporated or actively incorporating, have a live website and identifiable founding team, and be in production or planning to launch within six months. We consider larger funding rounds case by case. Full eligibility criteria are on the program page, and the program terms spell out the rest.

Once approved, create a LiveKit project and start building. If this is your first voice agent, our quickstart will get you up and running in about ten minutes.

There are incredible startups building the future of voice, video, and physical AI. We want to give them more room to build.

Apply now.