Usage, cost, and corrections
in a single view
See usage and costs across generation and delivery, along with how often Gateway repairs invalid output.
7% of UI generations fail silently in production
Every failed generation still returns 200. Based on 1,285 production failures observed over 15 days.
- 44%No valid root
- 36%Reference graph
- 16%Enum, type and
argument errors - 4%Truncation
OpenUI matches or beats Google A2UI and Vercel json-render on structural validity, with about half the tokens and streaming time. Better models reduce the error rate. None get it to zero
Fix invalid output before users see it
Every response is validated against your component library. Invalid ones are repaired in the streaming path. Over 99% render successfully.
No token markup
Tokens are billed at model providers’ standard rates.
OpenAI-compatible
Chat Completions and Responses endpoints work with your existing SDKs.
Bring your own API keys
Use your existing OpenAI, Anthropic, or Google Vertex credentials and commitments.
Works with other gateways
Run Gateway behind Portkey or any other OpenAI-compatible gateway.
Keep your cache setup
Your existing upstream cache configuration continues to work.
Broad model support
Use models listed on models.dev, with more available through OpenRouter.
Adopt it by changing your base URL
- 1Generate an API key
Create a Gateway key in the OpenUI console.
- 2Update your base URL and model
Use the Gateway endpoint as your base URL and choose any model string from models.dev.
- 3Enable Gateway in your prompt
Include your component schema so Gateway can validate each response.
const client = new OpenAI({ apiKey: process.env.THESYS_API_KEY, baseURL: "https://api.thesys.dev/v1/embed",});
const response = await client.chat.completions.create({ model: "openai/gpt-5", messages: [ { role: "system", content: generatePrompt({ thesys: true }) }, ...messages, ],});Built for enterprise requirements
View trust centreZero data retention
Chat Completions requests have zero data retention by default on paid models.
No training on your data
Your data is not used for training, except with free models.
Compliance
GDPR, SOC 2, and ISO 27001 details are available in the Trust centre.
Provider fallbacks
If a provider is unavailable, Gateway serves the same model through another provider.
Service uptime
View current and historical uptime on our service status page.
Private deployment
Self-hosting and VPC deployments are available on Scale.
Built to scale
with your usage
Each plan includes monthly API calls and all correction calls. Every response uses an API call plus LLM tokens. Tokens are billed separately at model provider rates.
Gateway plans. Annual billing selected; prices shown per month.
Free
$0- API calls/month
- 3K
- Overage
- —
- Includes
- Bring your own LLM key
Build
$49/mo$588 billed annually- API calls/month
- 25K
- Overage
- $0.002/call
- Includes
- Email support
Grow
$499/mo$5,988 billed annually- API calls/month
- 500K
- Overage
- $0.001/call
- Includes
- SSO/SAML, priority support
Scale
Custom- API calls/month
- Custom
- Overage
- —
- Includes
- Self-hosting and VPC, SLAs
Frequently asked questions
Have another question? Join our Discord.
Do I need OpenUI Gateway to use OpenUI?
No. OpenUI works without Gateway. You can call any supported LLM directly and use generatePrompt to create the system prompt.
For production applications, OpenUI Gateway adds validation and automatic corrections for generated UI, model routing, and provider fallbacks through an OpenAI-compatible API.
What happens to my data?
Your data is not used to train models, except with free models. Chat Completions requests have zero data retention by default on paid models.
Can I use my existing OpenAI or Anthropic credits?
Yes. Bring your own key (BYOK) to use your existing OpenAI or Anthropic credits and commitments. BYOK carries no platform fee.
I already use open-source OpenUI. How do I add Gateway?
Create a Gateway API key, update your base URL and model, and enable Gateway in your prompt. The integration guide covers the full setup.
You can also give the setup prompt to your coding agent or book a call for help.
Can I use OpenUI Gateway with Portkey or another gateway?
Yes. Configure OpenUI Gateway as an OpenAI-compatible endpoint behind Portkey or any other OpenAI-compatible gateway.
Will my cache configuration work?
Yes. OpenUI Gateway honors your upstream cache configuration.
Which model handles corrections, and are they billed separately?
Corrections use a dedicated model tuned for low latency, which we continue to improve. All correction calls are included in your plan and are not billed separately.
Do I need OpenUI Observability with Gateway?
No. OpenUI Gateway and OpenUI Observability work independently. Gateway does not require Observability.
We recommend using both. Gateway makes generated UI more reliable. Observability shows what users experienced and whether it worked.