What problem does it solve? Building apps where AI agents are first-class citizens requires a different architecture than traditional feature development, and most teams lack concrete patterns for tool design, action parity, context injection, and agent execution loops. ## Core Features & Use Cases - Architecture Patterns: Apply event-driven agents, unified orchestrators, two-layer git, and agent-to-UI communication patterns grounded in five core principles (parity, granularity, composability, emergent capability, improvement over time). - Tool & Prompt Design: Design atomic MCP tools with CRUD completeness, dynamic capability discovery, and system prompts that define features as outcomes rather than code. - Execution & Testing Guidance: Implement explicit completion signals, checkpoint/resume, model tier selection, and outcome-focused agent testing with parity audits. - Use Case: When adding a "publish to feed" UI feature, use the action parity workflow to add the corresponding agent tool, update the system prompt, and verify with a natural language test in the same PR. ## Quick Start Ask the agent to help design an agent-native architecture for your app, then pick a topic number from the intake menu such as tool design or action parity.
