/migrations

Update installation method: npm install replaced with direct binary download

npm install -g @anthropic-ai/claude-code

# Download from GitHub releases
curl -L https://github.com/anthropics/claude-code/releases/latest/download/claude-code-macos -o /usr/local/bin/claude-code
chmod +x /usr/local/bin/claude-code

// npm package deprecated in v2.0. Binary installation is now recommended.

Migrate settings.json structure: Permissions model changed to granular tool-based system

{
  "autoApprove": true,
  "permissions": "all"
}

{
  "permissions": {
    "allow": [
      "Read(*)",
      "Write(*)",
      "Bash(npm *)"
    ]
  }
}

// Old autoApprove flag removed. Use permission wildcards instead.

Update CLAUDE.md location: Project instructions moved from root to .claude/

./CLAUDE.md

./.claude/CLAUDE.md

// All Claude Code files now live in .claude/ directory

Add MCP tool configuration: Enable Model Context Protocol servers

# No MCP support

# .claude/settings.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/allowed"]
    }
  }
}

// MCP servers extend Claude with custom tools. Optional feature.

Migrate to new skills directory: Custom commands now use .claude/skills/

.claude/commands/my-command.md

.claude/skills/my-skill/SKILL.md

// Skills have dedicated directories for better organization

Enable architect mode: New mode for large refactoring tasks

aider --edit-format whole

aider --architect

// Architect mode uses planning before implementation

Configure model settings file: Persistent model configuration

aider --model gpt-4 --no-stream

# .aider.model.settings.yml
model: gpt-4
stream: false

// Reduces repetitive CLI flags

Update auto-commits flag: Flag renamed for clarity

aider --no-auto-commit

aider --no-auto-commits

// Plural form for consistency

Migrate diff format: Default diff format changed

# Used whole file diffs by default

aider --edit-format udiff

// udiff is now default. Use --edit-format whole for old behavior.

Update extension settings: Settings namespace changed

{
  "cline.apiKey": "...",
  "cline.model": "claude-3-opus"
}

{
  "cline.api.key": "...",
  "cline.api.model": "claude-sonnet-4-5"
}

// Settings now nested under api namespace

Configure custom instructions: New global instructions file

# Instructions in VSCode settings

# Create .cline/instructions.md in workspace root
# Project-specific instructions
Always use TypeScript strict mode
Prefer functional patterns

// Workspace-level instructions override global settings

Rename config file: Configuration moved to standard location

~/.codex/config.yaml

~/.config/codex/config.yaml

// Follows XDG Base Directory specification

Update API key environment variable: Standardized environment variable naming

export OPENAI_KEY=sk-...

export CODEX_API_KEY=sk-...

// Tool-specific env var to avoid conflicts

Migrate model selection syntax: New model specification format

codex --model gpt-4

codex --model openai:gpt-4

// Provider prefix now required for clarity

Enable Composer agent mode: New autonomous coding mode

# Standard chat mode only

# Open Composer (Cmd+Shift+I)
# Enable 'Agent' mode in Composer settings

// Agent mode allows multi-file edits with minimal supervision

Configure .cursorrules: Project-specific instructions

# No project instructions

# .cursorrules
Use Bun instead of npm
Prefer Tailwind utility classes
Always write tests for new features

// Cursor reads .cursorrules automatically on project load

Update model selection: GPT-4 Turbo deprecated

gpt-4-turbo

gpt-4o

// GPT-4o is faster and more capable than Turbo

Enable prompt caching: Reduce API costs with caching

gemini-cli chat

gemini-cli chat --cache

// Caching can reduce costs by 90% for repeated context

Configure model preference: Set default model in config

gemini-cli --model gemini-2.0-flash-exp

# ~/.gemini-cli/config.json
{
  "defaultModel": "gemini-2.5-pro"
}

// Gemini 2.5 Pro has better code generation capabilities