A practical lab for
learning Claude
by building real
experiments.
Eggthropic explores Claude Code, Agent Skills, MCP servers, API patterns, and UX experiments — all documented honestly: what works, what fails, and what comes next.
Four labs, one mission
Each lab focuses on a different layer of the Claude developer ecosystem — from CLI workflows to protocol-level integrations.
Claude Code Lab
Experiments using Claude Code — the agentic CLI that reads your codebase, runs commands, and ships committed code across files.
Agent Skills Lab
Building portable, reusable Agent Skills: SKILL.md directories that give Claude persistent capabilities across projects and platforms.
MCP Lab
Exploring the Model Context Protocol — building servers, clients, and interactive tools around the open JSON-RPC standard.
UX/UI Experiments
Prototyping AI-native interface patterns: streaming displays, tool-call visualizations, and interaction models beyond the chat bubble.
From the lab
Each experiment documents the goal, the prompt, what worked, and what failed.
Anthropic updates, decoded
Short articles on Claude and Anthropic developments: what changed, why it matters, what can be built.
Community Lab
Soon: a controlled public lab where contributors can propose Claude-powered experiments through GitHub pull requests.