yao: what it is, what problem it solves & why it's gaining traction

What it solves

Yao provides a unified runtime for building AI agents and web applications. It simplifies the development process by bundling everything needed—database models, REST APIs, a web UI, and an AI agent framework—into a single executable binary, removing the need for complex environment setups like Node.js or Python.

How it works

Yao operates as a controlled pipeline where every request passes through "Create Hooks" (pre-processing) and "Next Hooks" (post-processing). This allows developers to define strict boundaries and constraints around AI behavior. It supports three execution modes:

• LLM Mode: Uses models from providers like OpenAI and Anthropic for conversational tasks.
• CLI Agent Mode: Runs CLI runners (like Claude Code) in isolated containers with VNC desktop support for computer-use tasks.
• Pure Hook Mode: Uses TypeScript code for deterministic logic and routing without AI.

Who it’s for

Developers building AI-powered applications and agents that require a combination of deterministic logic, LLM-based reasoning, and isolated environment execution (sandboxing).

Highlights

• Single Binary Runtime: Includes a built-in V8 engine for TypeScript, data models, and server-side rendered UI pages.
• Agent Framework: Features native MCP support, a multi-agent delegation system, and a memory API with four distinct scopes (request, session, user, team).
• Integrated Search: Built-in support for Vector Search, Knowledge Graphs, and GraphRAG.
• Skills Ecosystem: Allows the use of reusable capability packs via SKILL.md files for CLI agents.