Deployment

Once your agent works locally, the next step is running it somewhere durable, reachable, and safe. This section covers how to take a Syntic Agent SDK program from a laptop prototype to a service your users and systems depend on. Amara, running on the Syntic model through api.syntic.ai, behaves the same in production as it does in development — but the environment around it changes considerably.

Deployment concerns fall into two broad categories. The first is operational: where the process runs, how it scales, how it manages concurrent sessions, and how it maintains a stable connection to the Syntic API. The second is protective: how you constrain what the agent can touch, how you store credentials, and how you contain the blast radius of a tool call that goes wrong.

What this section covers

The Hosting page walks through running the SDK inside long-lived servers, containers, and serverless functions. It explains how to structure a request-handling loop, keep the Syntic API connection warm, and manage sessions when many users share a single deployment.

The Secure Deployment page focuses on hardening. It covers sandboxing the workspace, applying least-privilege permissions, handling API keys and secrets, and controlling the network so an agent can only reach what it genuinely needs.

Before you deploy

Deployment amplifies whatever your agent already does. If it can edit files, in production it can edit production files; if it can run shell commands, it can run them on your infrastructure. Read Control & Governance first and settle your permission model before exposing an agent to real traffic. Treat the two pages here as the operational and security halves of the same launch checklist.