Agentry

Hi HN,Over Thanksgiving weekend I wanted to build an AI agent. As a design exercise, I wrote it as a set of React components. The component model made it easier to reason about the moving parts, composability was straightforward (e.g., reusing agents/tools), and hooks/state felt like a reasonable way to model behavior that changes over time.As I kept experimenting, I realized I was more interested in the framework (and it was a good excuse to explore React’s internals) than in the agent I originally planned, so I built this project instead.In this framework:- Agents are React components.- Agents have <Message>, <Tools>, … and even other agents as tools via <AgentTool>.- You can mount/unmount tools based on React state, and tools can set state to change what tools are available on the next turn.- There is a <Condition when="<natural language>" /> component that allows changing behavior based on the conversation history.- Built-in support for code execution, web search, memory, and MCP servers through Anthropic’s API.This is still a WIP project, but I’d love any feedback.---Repo: https://github.com/colinds/agentrySelected examples:- Hooks: https://github.com/colinds/agentry/blob/main/packages/exampl...- Subagents: https://github.com/colinds/agentry/blob/main/packages/exampl...- Conditions: https://github.com/colinds/agentry/blob/main/packages/exampl...