AI Agent Integration
The CLI provides optimized output for AI coding agents, enabling autonomous review-fix loops.Supported Agents
- Claude Code — via
--prompt-onlyoutput - Cursor — via
--prompt-onlyoutput - Windsurf — via
--prompt-onlyoutput
The Review-Fix Loop
The--prompt-only flag outputs minimal, structured analysis that AI agents can consume directly. This enables a powerful workflow:
Agent Runs Review
Team Key Authentication for Agents
For AI agents, use a team key so no individual login is needed:Decision Memory
Capture and persist the reasoning behind AI agent decisions directly in your repository.How It Works
When enabled, Kodus hooks into your AI agent’s turn-complete events and captures decisions as structured markdown files in.kody/.
Setup
Workflow
Work Normally
As you (or your AI agent) work on code, decisions are automatically captured on each turn-complete event.
Context Files
The CLI reads project context from these files if they exist in your repository:| File | Description |
|---|---|
.kodus.md | Kodus-specific configuration and guidelines |
claude.md | Claude-specific guidelines |
.cursorrules | Cursor IDE rules |