Beta
Govern every model call. Install nothing.
NEXORA Gateway is a hosted door in front of your model provider. Your apps call one endpoint with one key. Every request is authenticated, limited, and logged, and there is no package to install, patch, or trust inside your own stack.
curl https://api.onenexora.com/v1/chat \
-H "Authorization: Bearer nx_live_..." \
-H "Content-Type: application/json" \
-d '{"messages":[{"role":"user","content":"Hello"}]}'Why a hosted gateway
Self-hosted gateways become part of your attack surface.
An open-source proxy you run holds your provider keys and sees every prompt. When one of its releases is compromised, so is everything that trusted it. With Gateway there is nothing of ours inside your build.
You can't govern what you can't see.
Teams share provider keys and have no record of who sent what. Gateway gives every app its own key, limits, and an audit log, so you can see usage and shut a key off in one click.
Key. Call. Govern.
- 01
Key
Create a Gateway key in the dashboard. It is shown once and stored only as a hash.
- 02
Call
Send a chat request to one endpoint. Same shape whichever provider sits behind it.
- 03
Govern
Every call is authenticated, rate limited and checked against your quota. Calls that go through are written to your audit log.
What is real today
- One governed endpoint and one key in front of your provider
- Authentication, rate limiting, and monthly quota on every call
- Request logs with decision, latency, and token counts, and a prompt hash instead of prompt text
- Your own Anthropic or OpenAI key, checked and stored encrypted
- Model allow-lists, and checks for sensitive data and prompt injection, in monitor, mask or block mode
- A playground to test prompts without a key
On the roadmap
- Routing rules and fallback across providers
- More policies, such as per-key rules and spend limits
- Machine-learning detection and custom rules
- SIEM export and webhooks
Questions
- Do you store my prompts?
- No. We store a hash of each prompt with the model, token counts, latency, and decision. See Data handling.
- Which providers can I use?
- Anthropic and OpenAI, with your own key or the shared one. Routing rules across providers are on the roadmap.
- What does it cost?
- Free and Starter plans are live. See Pricing.
- Does it stream?
- Yes. Set stream to true and read the response as server-sent events.