SynOI

Built for whoever owns this deployment decision

Does the AI traffic stay on your network,
or can it go to ours?

That's the decision this page is for. SynOI Edge runs entirely on your machine: it's what production deployments actually run on today, for teams that need the traffic to never leave. SynOI Cloud, a SynOI-operated hosted gateway, is not live yet: request early access below and we'll notify you.

Either way, SynOI sits between your agents and the model calls they make: reusing cached answers, enforcing policy before a call fires, and routing to the cost tier you configure. One environment variable. No agent code changes.

Available now

SynOI Edge

Beta

The local agent, and what production deployments run on today. SynOI Edge runs entirely on your machine: cache, governance, and routing at wire speed, with no cloud dependency. Built for latency-critical and offline-first deployments.

What it does

  • →Semantic cache: reuse answers across repeated and similar requests, local, no network round-trip
  • →Runs offline: works without internet access to SynOI
  • →NPU / GPU / SmartNIC acceleration: hardware-offloaded OID lookup on supported hardware
  • →Sentinel policy, signed audit trail: same governance logic as the rest of the gateway
  • →On-device only: data never leaves the machine
  • →Runs in-process: library or eBPF probe, your choice
synoi-edge.ts
// One env var. Drop-in replacement.
const client = new SynOI({
  apiKey: process.env.SYNOI_KEY,
  mode: 'edge',
})

Beta: Functional today, still in flux. Self-host on your own hardware, forever, for free; no upfront commit.

Start with the Quickstart

docker compose up. Working gateway in five minutes.

Coming soon

SynOI Cloud

Not live yet

A SynOI-operated hosted gateway: the same governance and routing logic as Edge, without running it yourself. Not live yet: production deployments today run on SynOI Edge (self-hosted).

What it will do

  • →Semantic cache: reuse answers across repeated and similar requests
  • →Named routing presets: point a call at a cost tier you configure
  • →Multi-provider: Anthropic, OpenAI, Google, and more behind one key
  • →Sentinel policy: pre-execution enforcement before any model call fires
  • →Signed audit trail: every decision receipted and replayable
  • →Context trimming: strip redundant tokens before they reach the model

Not yet available: SynOI Cloud is not live in production. Join the waitlist to get notified when hosted access opens; no payment required to join.

Join the Cloud waitlist

We'll notify you at launch.

Enterprise

Regulated industries.
Custom deployments.

Air-gapped deployment, custom encoder fine-tuning, SDK-side tenant-encrypted payload mode, SSO, data processing agreements, and a dedicated engineer. We scope it to your stack and your compliance requirements.

Set an appointment →

We respond within one business day to schedule a scoping call.