multi-agent-orchestration

Multi-agent orchestration coordinates Codex for code and Gemini for UI from a Kiro spec, centralizing dispatch, review, and state synchronization to deliver features end-to-end. It uses AGENT_STATE.json and PROJECT_PULSE.md to track progress and blockers.

Coordinating code and UI generation across agents uses a centralized orchestration model that dispatches tasks from a Kiro spec to Codex and Gemini. It manages auto-dispatch, review, and fix loops to synchronize state across both agents.

To start orchestration from a Kiro spec, initialize the provided tooling and trigger the workflow using the command Start orchestration from spec at <path>. This dispatches end-to-end feature tasks to Codex and Gemini agents.

Multi-agent orchestration supports deterministic and LLMS-driven modes that include auto-dispatch, review, fix loops, and pulse updates. This ensures tasks dispatched to Codex and Gemini are synchronized and reviewed automatically.

You can run orchestration for a specific feature by triggering the workflow with the command Run orchestration for <feature>. This coordinates the end-to-end feature work across Codex and Gemini agents based on your Kiro spec.

Multi-agent orchestration maintains coordinated state using AGENT_STATE.json and PROJECT_PULSE.md files. These files reflect decisions, progress, and blockers, ensuring synchronized state across dispatched Codex and Gemini workstreams.