Layercode FAQ

Dropping a bunch of frequently asked questions and answers here:

How does Layercode work?

Layercode uses a chained ‘voice pipeline’ approach to give LLM-powered AI agents the ability to listen to a user’s speech, transcribe and process the input, and respond with speech.

Here’s how Layercode works:

1. Capture: A user’s speech is captured, and audio streams to nearest edge location.

2. Transcribe: Convert speech to text using the industry leading speech-to-text transcription models.

3. Process: Your backend agent logic processes the text and sends a text response back to Layercode.

4. Synthesize: Layercode uses your chosen text-to-speech model to convert the agent’s response to natural speech.

5. Deliver: The speech is streamed back to the user.

Unlike realtime audio models (e.g. ChatGPT Realtime voice API) the pipeline approach gives developers full control over every component needed to process a user’s speech and generate a conversational response. This is far more reliable, easier to debug and scale, and cheaper.

Who is Layercode for?

Developers who want production-ready voice AI without months of infrastructure work or the limitations of no-code platforms.

Whether you're adding voice to an existing agent or building from scratch, Layercode handles the complex real-time infrastructure so you can focus on your engineering resources on your agent's intelligence and core product experience.

We’re building Layercode for:

1. Voice AI Startups building vertical-specific voice agents, who value fast-iteration, model flexibility, and true usage-based costs.

2. AI Engineers at Scale-ups building customer-facing voice features for existing products, looking for production reliability, full backend control, and reliable global performance.

3. Enterprise Innovation Teams building internal productivity tools and POCs, who need to meet advanced security and compliance requirements.

How is Layercode different from other voice AI platforms?

Layercode occupies a unique position between no-code platforms (Vapi, Retell) and DIY solutions (LiveKit, custom WebRTC). We are also the first true edge-native voice AI platform, leveraging a 330-location global edge network.

Layercode’s main benefits to developers:

1. Fast, easy setup via the command line: We make it easy to get started with a single CLI command. Connect any LLM, agent framework, or custom logic via a simple webhook

2. Full control of your agent’s backend: Developers looking to build complex, finely tuned AI applications need full control over their agent’s backend logic:

   1. Serious developers will quickly hit the limits of the visual workflow builders offered by no-code agent platforms, which also don’t support complete backend control.

   2. Some developers may prefer to invest significant time and resources into building and maintaining every component of their voice pipeline. But, when agents hit production, this comes at a continued cost.

   3. We believe the majority of voice AI use cases will benefit hugely from leveraging a scalable platform by allowing developers to focus their resources on developing their core products.

3. True, usage-based pricing: We charge per-second. You only pay for active speech — silence is always free. This reduces the actual per-minute cost to run your agent in production.

4. Deliver consistently reliable low-latency voice AI to end-users: Production voice agents that run on ****competitors’ regional cloud infrastructure often suffer from unreliable latency and performance issues. Layercode’s 330+ location global edge network helps ensure:

   1. Zero cold-start times ensure there’s never a delay to the start of a conversation.

   2. Isolated call sessions avoid the resource contention that can lead to degraded performance and the ‘noisy neighbour’ problem that occurs when voice AI platforms run multiple concurrent sessions on shared infrastructure.

   3. Global reliability means users get the same reliable conversational experience, wherever they may be in the world.

How do I get started with Layercode?

Sign up for a free Developer account and start building with $100 credits (enough for ~1,000 minutes of agent conversation).

You can use Layercode CLI to get started right from your terminal. Just run:

npx @layercode/cli init

Then, authenticate with your Layercode account. Choose a template, and our CLI sets up your tunnel and webhook URLs for local development.

Read our quickstart guide and documentation to start building.

How much does Layercode cost?

Layercode bills per second of conversation when the user or agent is speaking.

You only pay for what you use, in per-second increments. Silence is always free.

Your cost per-second is determined by the providers and models you choose for the transcription and text-to-speech stages of your voice pipeline. Check out our pricing calculator to estimate your usage →

If you're building voice AI at a startup, you might qualify for $2,000 free credits via our Startup Program.