Claude Code Overview
Architecture, philosophy, and capabilities
Claude Code is Anthropic's terminal-based coding agent, released in February 2025. It represents a deliberate architectural choice: rather than embedding AI into an IDE, Claude Code meets developers in the terminal the interface that works across every development environment.
The design follows Unix philosophy. Claude Code is composable and scriptable, designed to integrate with existing workflows rather than replace them.
A command like tail -f app.log | claude -p "alert me if you see errors" demonstrates this principle: the agent becomes one component in a larger pipeline.
What Claude Code can do
- Read files Access files anywhere in the codebase
- Write and edit Modify files directly, including multi-file coordinated changes
- Execute commands Run shell commands in a persistent bash session
- Search Use glob patterns and regex across the entire project
These capabilities combine into an autonomous loop: Claude reads context, plans actions, executes them, observes results, and iterates until the task completes. The developer directs and reviews; the agent implements and verifies.
Context and memory
Claude Code uses a 200,000-token context window roughly 500 pages of text.
Project-specific knowledge persists through CLAUDE.md files, which configure behavior, document conventions, and encode institutional knowledge. These files load automatically, providing agents with the context that would otherwise require repeated explanation.
MCP: Extending capabilities
The Model Context Protocol (MCP) allows Claude Code to connect to external systems:
- Databases
- APIs
- Issue trackers
- Custom tools
MCP transforms Claude Code from a codebase-bound agent into an integration platform, enabling workflows that span code, data, and external services.
Module 10 covers MCP configuration in detail.