Routing through an LLM gateway

Many organizations sit an LLM gateway between their developer tools and the model provider so that every request passes through a single, observable choke point. Syntic Code supports this pattern: instead of talking to api.syntic.ai directly, each syntic instance can be pointed at your gateway, which then forwards traffic to the Syntic model on your behalf. This lets a central team enforce rate limits, attach cost-center tags, redact sensitive fields, and keep one audit log for all model usage.

When a gateway helps

A gateway earns its place once you have more than a handful of users. It gives you a place to rotate credentials without touching every developer’s machine, to apply per-team quotas, and to collect uniform telemetry across editors, CI jobs, and the CLI. Because the gateway holds the real upstream key, individual developers never need to see it, which shrinks the surface area for credential leaks. Gateways are also the natural home for content filtering and prompt logging when compliance requires them.

Pointing Syntic Code at your gateway

Configure the base URL through an environment variable so that syntic sends requests to your proxy rather than the default endpoint:

export SYNTIC_BASE_URL="https://gateway.internal.example.com/syntic"
export SYNTIC_API_KEY="issued-by-your-gateway"

Your gateway must expose an API compatible with api.syntic.ai and translate the paths and headers Syntic Code sends. Return the same response shape, including streaming events, so the CLI renders replies incrementally. Set these variables through server-managed settings or a device management profile so they apply automatically and cannot be pointed elsewhere.

Operational considerations

Keep the gateway’s added latency low, since Amara streams tokens interactively and users feel any delay. Preserve streaming end to end rather than buffering whole responses. Finally, monitor the gateway itself: if it becomes unavailable, every Syntic Code session that depends on it stops, so treat it as production infrastructure with health checks and failover.