ngrok launches AI Gateway: one URL to route cloud and self-hosted models

ngrok has launched AI Gateway, a managed gateway that unifies cloud LLM APIs and self-hosted models behind a single URL, with the tagline "Never bet on one model." Multiple posts (m1–m7) relayed the same announcement with consistent details.

Confirmed

• Minimal integration: change the SDK baseURL to gateway.ngrok.ai and use an ngrok API key; works with OpenAI/Anthropic/Vercel AI SDK tooling, no deployment, VPN, or per-provider integration needed.
• Model switching via the model parameter: local models and cloud models like OpenAI and Anthropic are callable through the same gateway.
• Local model scenarios: per @smohinii's thread, Ollama on a laptop, vLLM on a server, and fine-tuned models behind firewalls can all connect without public exposure or tunnels, avoiding separate APIs, keys, SDKs, and billing accounts per model.
• Built-in capabilities: the gateway includes observability and access control; demos showed local and cloud models accessed through one gateway.

Why it matters

Switching between local and multiple cloud providers previously meant separate integrations, auth, and networking. AI Gateway flattens these differences into one URL, lowering the cost of multi-model routing, migration, and hybrid (local + cloud) deployment—especially attractive for teams that want flexibility across vendors without lock-in.

2026-09-03 ~ 2026-09-05 · 7 related posts

• [source] ngrok launches AI Gateway: route local LLMs and OpenAI through one URL — anthara_ai · 2026-09-03
• ngrok Launches AI Gateway: One URL to Route Cloud and Local Models — anthara_ai · 2026-09-04
• [source] ngrok AI gateway: one URL for cloud and firewall-bound local models — s_mohinii · 2026-09-04

4 near-duplicate retellings: anthara_ai · s_mohinii · anthara_ai · anthara_ai