Realtime TTS-2 is live. Built for realtime conversation that feels human. Read the Realtime TTS-2 announcement

Go live

Go to production

What to set up before you launch on Inworld: credentials, rate and concurrency limits, error handling, usage, billing, data handling, and team access.

Once your first requests work, use this page to prepare for launch. It covers what every Inworld API shares, and points to the product guides for the rest.

Secure your credentials

  • Authentication overview: pick the right credential for a server, a browser, a mobile app, or a Realtime client.
  • API keys: create keys with only the permissions your application needs.
  • Security best practices: never ship an API key in client code. Proxy through your backend, or mint a short-lived token for the client.

Plan for limits

Three limits apply, and they measure different things:

LimitWhat it capsGuide
Rate limitsHow often you can send requests.Rate limits
Concurrency limitsHow many requests or connections run at the same time.Concurrency limits
Job concurrency limitsHow many async and batch TTS jobs run at the same time.Job concurrency limits

Limits depend on your plan. See limits by product for where each product's numbers are listed.

A request over a rate limit returns 429. Retry it with exponential backoff and jitter instead of retrying immediately.

Handle errors and build for quality

Each API reports errors in its own format, so follow the guide for the product you use:

ProductGuides
TTSBest practices, latency, and concurrency errors
STTErrors and troubleshooting and best practices
LLM RouterRouting concepts for fallbacks, and streaming errors
Realtime APIError handling and tool calling

Track usage and cost

  • Usage: see consumption by product.
  • Billing: plans, features and limits by plan, and payment settings.

Meet data and access requirements

Get help

See Support for how to reach the team, what experimental, preview, and stable mean, and account questions.