Syntic Code SDK
Coming soon. A Syntic Code SDK for building your own agents on the same engine as the CLI is in the works. There is no published package yet, so there is nothing to install or set up today.
The SDK will let your own programs run Syntic Code sessions: send a prompt, receive the agent’s messages as they stream in, and answer permission questions in code. This page will hold the guides and the API reference once the first package is released.
What you can use today
You can already drive Syntic Code from scripts and other programs through the syntic command itself:
| Need | Use |
|---|---|
| Run one task and read the answer | syntic -p "<prompt>" prints the final reply and exits. |
| Machine-readable results | --output-format json returns one JSON object; --output-format stream-json emits one JSON event per line as the run goes. |
| Feed a conversation over standard input | --input-format stream-json reads user turns as JSON lines. |
| Answers that fit a shape | --json-schema asks for output that matches a JSON Schema. |
| Keep a run bounded | Limit tools with --allowed-tools and --disallowed-tools, and pick a permission mode with --permission-mode. |
See Scripting and CI for full examples, and the CLI reference for every flag.
Stay up to date
New releases and their notes are listed under What’s new.