Repository reasoning
Inspect source code, configuration, workspace context, and project structure before deciding how to proceed.
AI RUNNERFeatures
From intent to verified execution.
AI Runner combines repository reasoning, controlled execution, approvals, and verification in one autonomous engineering workflow.
Autonomous workflow
Reason. Act. Verify.
AI Runner works through engineering objectives while keeping execution observable and important changes under explicit control.
Controlled execution
Execute terminal operations and controlled changes through deterministic runtime boundaries.
Verification loops
Continue from execution into verification so outcomes can be checked against the engineering objective.
Stay in control
Autonomy without giving up oversight.
Explicit checkpoints
Review controlled mutations individually, approve sequentially, reject a proposed action, or stop execution.
Auto-Approve
Reduce repeated manual approvals when appropriate while keeping execution inside AI Runner's runtime controls.
Provider choice
Work with supported models from OpenAI, Anthropic, and Google through a unified AI Runner workflow.
AI Runner