An Apache-2.0 TypeScript framework for building AI agents — workflows, memory, RAG and evals — with a local studio and an agentic software factory on top.
ngrok AI Gateway
Summary
A hosted LLM gateway from ngrok: one base URL and one key for public providers, your own keys and self-hosted models, with failover, scoped access and cost analytics.
Screenshots
Description
ngrok AI Gateway is a hosted routing layer for language-model traffic, built by the team behind the ngrok tunnelling service. The integration is deliberately small: point your SDK's baseURL at https://gateway.ngrok.ai, swap in an ngrok access key, and every call your application already makes starts flowing through the gateway.
What the gateway adds
- One endpoint, many providers. Route to OpenAI, Anthropic, Google, Groq, DeepSeek, OpenRouter and others through a single URL, using the OpenAI, Anthropic or Vercel AI SDKs unchanged.
- Self-hosted and local models. Models you run yourself - Ollama, vLLM or a GGUF build on your own hardware - can be reached over ngrok's private connectivity, with no public IP or inbound port to open.
- Bring your own keys. Drop in the provider keys you already pay for and keep your existing rates; you continue to be billed by the provider directly, but every key is managed in one place. You can also start with ngrok credits and no provider account at all.
- Scoped access control. Issue a separate access key per app or per developer and set exactly which providers and models each one may call, instead of circulating a single key that opens everything.
- Failover and retries. Declare fallback models in the request, and when a provider or key degrades the gateway reroutes to a healthy alternative and retries failed requests without any error-handling code on your side.
- Observability. Tokens, latency, errors and spend are rolled up across every call and attributed to the app, developer and model that produced them - the breakdown provider dashboards do not give you.
Who it is for
Teams shipping LLM features who have outgrown a single hard-coded provider: anyone juggling several vendors, mixing hosted and self-hosted inference, or needing to attribute AI spend per team or per customer.
Things to know
Usage runs on prepaid ngrok.ai credits with a minimum $5 top-up, and access keys are created from the ngrok dashboard. Documentation for the gateway lives in the main ngrok docs.
Reviews
Similar App Suggestions
AI design engineer that generates distinctive UI designs and production code inside your own repo, Figma and design system.
Menu-bar LLM inference server for Apple Silicon, with continuous batching and tiered KV caching that keeps local models fast enough for real coding work.
Open-source local inference server that profiles your hardware, picks models that fit, and points your coding agent at them — free, private and offline.
Open-source AI coding agent for VS Code, JetBrains and the terminal — 500+ models at provider cost, task-specific modes, parallel agents in isolated git worktrees.