NativePort
Every web API behind one key — for developers and their agents
NativePort is an API gateway: create one account, hold one key, and reach 30+ search, scraping, browser-automation, voice, document-parsing and model-inference APIs through their own native interfaces. There are no per-provider accounts to open and no key ring to babysit — the gateway keeps each upstream credential server-side and attaches it to your requests for you.
Four of those — OpenAI, Anthropic, Grok and Hugging Face — are policed rather than transparent: the gateway enforces a model allowlist and a request-field policy on them, rejecting an unrecognized model or field before the call ever reaches the upstream. Browserless, Steel, Mistral and LlamaParse are policed too, in a narrower way — closed field policies govern their fixed actions, with no model catalog behind any of them. See each page below for specifics. A fifth route, Inference API, unifies the four model-inference routes behind one OpenAI-compatible chat-completions contract addressed by a canonical model id.
Why route through it
Existing integrations keep working: requests forward to each provider’s real endpoints, and responses come back untouched.
A single credit balance meters the whole catalog. No stack of invoices, no per-service billing cycles, no idle subscriptions.
The moment your key exists, every provider in the catalog answers it — no onboarding queues, provider by provider.
Upstream credentials never enter your code or your agent’s context; the gateway injects them per request.
Tools
Every proxied upstream, its route prefix, and the methods it accepts:
How it works
Create your key
Sign up — a single key is minted automatically on first sign-in, and it covers the entire catalog.