Syntic Agent SDK
The Syntic Agent SDK lets you build production agents on the same engine that powers Syntic Code, our agentic coding CLI. It ships for both TypeScript and Python, connects to Amara — our AI running on the Syntic model through api.syntic.ai — and gives you programmatic control over the full agent loop: prompting, tool use, permissions, sessions, and streaming.
Where the Syntic Code CLI is an interactive terminal experience, the SDK exposes the machinery underneath so you can embed agents inside your own applications, services, and automation. Anything the CLI can do — read and edit files, run commands, call custom tools, manage subagents — you can drive from code.
What you can build
The SDK is designed for a wide range of agentic workloads. Teams use it to power in-product coding assistants, autonomous refactoring pipelines, CI bots that triage and fix failing builds, and internal tooling that reasons over private codebases. Because the agent loop, tool surface, and permission model are all programmable, you decide how much autonomy Amara has and where a human stays in the loop.
The SDK is transport-agnostic about how you feed it work. You can send a single prompt and await a result, stream a live conversation, or wire the agent into a long-running service that persists sessions across requests.
How this documentation is organized
Start with the Quickstart to install the SDK and run your first agent. From there:
- Core Concepts explains the agent loop, sessions, and Syntic-specific features.
- Input & Output covers streaming, approvals, and structured results.
- Tools shows how to give Amara custom tools, connect MCP servers, and delegate to subagents.
- Customization covers system prompts, slash commands, skills, and plugins.
- Control & Governance covers permissions, hooks, checkpointing, cost tracking, and observability.
- Deployment covers hosting and hardening.
- API Reference documents the TypeScript and Python surfaces and migration.