How Syntic Code Works

Syntic Code is an agent, not a chat box. When you give Amara a task, it runs a loop: gather context, plan, act, and review — repeating until the work is done or it needs your input.

The agent loop

Every session follows the same rhythm. First Amara reads the parts of your repository that matter to the request, using search and file inspection rather than loading everything. Next it plans, breaking the task into concrete steps and, for anything non-trivial, summarizing the approach so you can approve it. Then it acts, editing files and running commands. Finally it reviews the outcome — reading test output, checking diffs, and correcting course when something fails. The loop continues automatically, so a single request can produce several coordinated changes.

Grounded in your repository

Amara works from what is actually in your codebase, not from assumptions. It searches for the relevant symbols, reads neighboring code to match your conventions, and respects the rules and memory you have defined in the .syntic/ directory. Because it operates inside your working tree, every edit is visible in your version control and easy to roll back.

You stay in control

The loop is designed around consent. Amara asks before running commands that could change your system, and you can approve, reject, or refine any step. You can interrupt at any time to redirect the work, and you decide when the task is complete. This keeps the speed of automation while leaving judgment where it belongs — with you. To go deeper, see how the context window shapes what Amara can consider on each turn.