Dit artikel beschrijft de ontwikkeling van Claude Code, verdeeld over vier belangrijke fasen:
De Basis (0.2.x → 1.0.0): De introductie van de CLI, Vim-bindings, het Model Context Protocol (MCP) en 'Thinking' modes.
Stabiliteit (1.0.1 → 1.0.126): Integratie van Sonnet 4 en Opus 4, de lancering van TypeScript/Python SDK's en de introductie van custom Hooks.
De 'Agentic' sprong (2.0.0 → 2.0.76): De toevoeging van sub-agents, de /rewind functie om wijzigingen ongedaan te maken en een native VS Code extensie.
Professionalisering (2.1.0 → 2.1.5): Ondersteuning voor achtergrondtaken (Ctrl+B), integratie van Opus 4.5 en optimalisaties voor enterprise-omgevingen.
Daarnaast worden technische verbeteringen genoemd, zoals het gebruik van Rust voor snellere bestandzoekopdrachten en het oplossen van geheugenlekken in WASM en Tree-sitter.
Based on these notes, here is a thematic summary of the tool's development and its most significant features.
---
🚀 Major Milestones
1. The Foundation (Versions 0.2.x → 1.0.0)
Core Interaction: Introduced the CLI interface, Vim bindings, and image support (pasting images into the prompt).
MCP Integration: A massive focus on the Model Context Protocol (MCP), allowing Claude to connect to external servers and tools.
Thinking Mode: Introduced "Thinking" and "Ultrathink" modes, allowing the model to process complex logic before responding.
Context Management: Introduced @-mentions for files and CLAUDE.md for project-level memory.
2. General Availability & Stability (Versions 1.0.1 → 1.0.126)
Model Upgrades: Launched with Sonnet 4 and Opus 4.
Customization: Added Hooks (SessionStart, PreToolUse, etc.), allowing users to run custom scripts when certain events occur.
Refined Tools: Redesigned the Search (Grep) tool and improved the Bash tool for better security and reliability.
SDK Release: Released TypeScript and Python SDKs to allow developers to build on top of the agent.
3. The "Agentic" Leap (Version 2.0.0 → 2.0.76)
Sub-agents: Introduced specialized agents (e.g., the Explore agent for codebase navigation) and the ability for Claude to spawn sub-agents for complex tasks.
The "Rewind" Feature: Added /rewind, allowing users to undo code changes—essentially a "git undo" for AI edits.
IDE Integration: Launched a native VS Code extension, bringing the CLI experience directly into the editor.
Plan Mode: Enhanced the "Plan" workflow, where Claude builds a precise execution strategy before writing code.
Background Tasks: Added the ability to run agents and bash commands in the background (Ctrl+B), allowing the user to keep working while Claude processes a long task.
Advanced Model Support: Integration of Opus 4.5.
Enterprise Readiness: Added environment variable overrides for temp directories, background task disabling, and managed settings for corporate environments.
---
🛠 Key Feature Summary
Category
Notable Features
Developer Experience
/rewind (undo), /context (debug context), /stats (usage tracking), Plan Mode.
Native Windows support, WSL support, CLAUDE.md for project memory.
📉 Model Evolution Path
The tool has tracked the release of Anthropic's most powerful models: Sonnet 3.7 → Sonnet 4 → Opus 4 → Haiku 4.5 → Opus 4.5.
⚠️ Notable Technical Fixes
Memory: Fixed several WASM and Tree-sitter memory leaks that occurred during long sessions.
Performance: Implemented native Rust-based fuzzy finders for faster file suggestions in large repositories.
Compatibility: Significant work on terminal rendering to support iTerm2, Kitty, Ghostty, and WezTerm.
This is a comprehensive changelog for Claude Code, documenting its evolution from an early beta (0.2.x) to its current stable versions (2.1.x).
Based on these notes, here is a thematic summary of the tool's development and its most significant features.
---
🚀 Major Milestones
1. The Foundation (Versions 0.2.x → 1.0.0)
Core Interaction: Introduced the CLI interface, Vim bindings, and image support (pasting images into the prompt).
MCP Integration: A massive focus on the Model Context Protocol (MCP), allowing Claude to connect to external servers and tools.
Thinking Mode: Introduced "Thinking" and "Ultrathink" modes, allowing the model to process complex logic before responding.
Context Management: Introduced @-mentions for files and CLAUDE.md for project-level memory.
2. General Availability & Stability (Versions 1.0.1 → 1.0.126)
Model Upgrades: Launched with Sonnet 4 and Opus 4.
Customization: Added Hooks (SessionStart, PreToolUse, etc.), allowing users to run custom scripts when certain events occur.
Refined Tools: Redesigned the Search (Grep) tool and improved the Bash tool for better security and reliability.
SDK Release: Released TypeScript and Python SDKs to allow developers to build on top of the agent.
3. The "Agentic" Leap (Version 2.0.0 → 2.0.76)
Sub-agents: Introduced specialized agents (e.g., the Explore agent for codebase navigation) and the ability for Claude to spawn sub-agents for complex tasks.
The "Rewind" Feature: Added /rewind, allowing users to undo code changes—essentially a "git undo" for AI edits.
IDE Integration: Launched a native VS Code extension, bringing the CLI experience directly into the editor.
Plan Mode: Enhanced the "Plan" workflow, where Claude builds a precise execution strategy before writing code.
Background Tasks: Added the ability to run agents and bash commands in the background (Ctrl+B), allowing the user to keep working while Claude processes a long task.
Advanced Model Support: Integration of Opus 4.5.
Enterprise Readiness: Added environment variable overrides for temp directories, background task disabling, and managed settings for corporate environments.
---
🛠 Key Feature Summary
Category
Notable Features
Developer Experience
/rewind (undo), /context (debug context), /stats (usage tracking), Plan Mode.
Native Windows support, WSL support, CLAUDE.md for project memory.
📉 Model Evolution Path
The tool has tracked the release of Anthropic's most powerful models: Sonnet 3.7 → Sonnet 4 → Opus 4 → Haiku 4.5 → Opus 4.5.
⚠️ Notable Technical Fixes
Memory: Fixed several WASM and Tree-sitter memory leaks that occurred during long sessions.
Performance: Implemented native Rust-based fuzzy finders for faster file suggestions in large repositories.
Compatibility: Significant work on terminal rendering to support iTerm2, Kitty, Ghostty, and WezTerm.
Zichtbare categorieën
Kies welke categorieën je wilt zien.
Over deze site
Waarom Tech-update?
Tech-update is ontstaan omdat ik merkte dat ik dagelijks veel verschillende websites moest bezoeken om interessant technieuws te vinden. Dat kon efficiënter.
Hoe werkt het?
Meerdere keren per etmaal gaat mijn zelfgebouwde AI-agent op zoek naar interessante artikelen op blogs en nieuwssites.
Vindt hij iets relevants, dan wordt het artikel naar het Nederlands vertaald en op Tech-Update geplaatst.
Van ieder artikel maakt de AI ook een korte samenvatting. Zo kun je snel bepalen of het de moeite waard is om het volledige artikel te lezen.
Bij ieder artikel staat altijd een link naar de oorspronkelijke bron (vaak een Engelstalige website).
Artikelen kun je bewaren als favoriet, vastzetten in de zijbalk of terugvinden via de zoekfunctie.
In de instellingen (tandwielikoon rechtsboven) geef je aan in welk soort nieuws je geïnteresseert bent.