Tool Migration Checklist
Migrating to or between Cursor, Claude Code, and Codex follows four stages: assess, pilot, migrate, optimize. A team rollout runs a pilot in weeks 1-2, early adopters in weeks 3-4, full rollout in weeks 5-8, and optimization from month 3, carrying configuration files, success metrics, and a rollback plan through each stage.
You have decided to adopt a new AI coding tool — or switch from one to another. This checklist walks you through every step so nothing falls through the cracks. Whether you are moving a single developer or an entire team, the process is the same: assess, pilot, migrate, optimize.
What this migration checklist covers
Section titled “What this migration checklist covers”- A step-by-step migration plan with checkboxes you can track
- Feature mapping tables between tools
- Team rollout timelines with realistic milestones
- Common pitfalls and how to avoid them
Pre-Migration Assessment
Section titled “Pre-Migration Assessment”Before touching any tool, answer these questions:
- ☐ What tools are you migrating FROM? (VS Code, Copilot, ChatGPT, another AI tool)
- ☐ What tools are you migrating TO? (Cursor, Claude Code, Codex, or a combination)
- ☐ How many developers are affected?
- ☐ What is your budget per developer per month?
- ☐ Do you have compliance requirements? (SOC 2, GDPR, HIPAA)
- ☐ What is your acceptable productivity dip during transition? (1 week? 2 weeks?)
- ☐ Who are your pilot users? (2-3 early adopters)
- ☐ What does success look like? (Define metrics upfront)
Migration Paths
Section titled “Migration Paths”From VS Code
Section titled “From VS Code”-
One-click import: Open Cursor > Settings (
Cmd/Ctrl+Shift+J) > General > Account > “Import” under VS Code Import. Transfers extensions, themes, settings, and keybindings. -
Verify extensions loaded: Some extensions may need manual reinstallation. Check View > Extensions.
-
Cursor-specific setup:
- Create
.cursor/rules/directory with project guidelines - Configure AI model preferences (Settings > Models)
- Enable Privacy Mode if required (Settings > Privacy)
- Install essential MCP servers (Settings > Features > MCP)
- Create
-
Learn the new shortcuts:
Cmd/Ctrl+I(Agent),Cmd/Ctrl+K(Inline Edit),Tab(autocomplete),@(context references)
From JetBrains IDEs
Section titled “From JetBrains IDEs”-
Install keybinding extension: Search for “IntelliJ IDEA Keybindings” in Extensions
-
Recreate core features: Install language-specific extensions, configure debuggers, set up database tools
-
Key differences:
- File navigation:
Ctrl+Pinstead of double Shift - Refactoring: Different shortcuts, similar capabilities
- Built-in tools: Some JetBrains features need extensions
- File navigation:
From GitHub Copilot
Section titled “From GitHub Copilot”| Copilot Feature | Cursor Equivalent |
|---|---|
| Ghost text suggestions | Tab autocomplete (faster, multi-line) |
| Copilot Chat | Agent mode (Cmd/Ctrl+I) |
| Comment-driven generation | Natural language in chat |
| Explain code | Ask mode |
| Fix suggestions | Inline edit (Cmd/Ctrl+K) |
| PR reviews | BugBot |
From Any IDE/Tool
Section titled “From Any IDE/Tool”-
Install Claude Code:
Terminal window curl -fsSL https://claude.ai/install.sh | bashclaude --version -
Initialize project: Navigate to project root and run:
Terminal window claude/init # Creates CLAUDE.md with project context -
Write comprehensive CLAUDE.md: Include coding standards, common commands, architecture notes, and test patterns.
-
Set up MCP servers:
Terminal window claude mcp add context7 -- npx -y @upstash/context7-mcp -
Configure permissions: Run
/permissionsto set which operations Claude can perform without asking. -
Install VS Code extension (optional): Search “Claude Code” in Extensions marketplace.
From Cursor
Section titled “From Cursor”| Cursor Feature | Claude Code Equivalent |
|---|---|
| Agent mode | Default behavior (Claude is always agentic) |
| Ask mode | Just ask a question (Claude reads files on demand) |
@ file references | @file.js in prompts |
.cursor/rules/ | CLAUDE.md + .claude/rules/ |
| Tab autocomplete | Use Cursor alongside for this |
Inline edit (Cmd/Ctrl+K) | Describe the edit in the prompt |
| Checkpoints | Native checkpoints and /rewind |
| Background Agent | Background agents in terminal/IDE, isolated worktrees, desktop/web tasks, or dynamic workflows |
From Any Tool
Section titled “From Any Tool”-
Install Codex CLI:
Terminal window npm install -g @openai/codexcodex --version -
Configure Codex:
Terminal window # Start Codex and use /init inside the TUI to set up configcodexThen run
/initinside the interactive session. Authenticate separately with ChatGPT orcodex login --with-api-key; keep credentials out of~/.codex/config.toml, which is for model, sandbox, and approval settings. -
Create AGENTS.md: Write project-level configuration with coding standards, test commands, and deployment patterns.
-
Choose your surface: Codex in the ChatGPT desktop app, CLI (terminal), IDE extension, or Cloud (background agents). Desktop tasks can use worktrees when you choose an isolated task; worktrees are not universal across every surface.
-
Set up automations (optional):
- Connect GitHub repository
- Configure Slack integration
- Set up Linear ticket automation
From Cursor / Claude Code
Section titled “From Cursor / Claude Code”| Previous Feature | Codex Equivalent |
|---|---|
| Agent mode / autonomous Claude | on-request plus workspace-write for interactive work; never plus an explicit sandbox only for trusted unattended jobs |
.cursor/rules/ / CLAUDE.md | AGENTS.md |
| Background Agent / headless mode | Codex Cloud agents |
| MCP servers | MCP servers (same protocol, same setup) |
| Custom commands / slash commands | Codex slash commands |
| BugBot | Codex GitHub automation |
Team Migration Strategy
Section titled “Team Migration Strategy”Phase 1: Pilot (Week 1-2)
Section titled “Phase 1: Pilot (Week 1-2)”- ☐ Select 2-3 pilot users (different skill levels, different project types)
- ☐ Document current workflows and pain points
- ☐ Install and configure the new tool for pilot users
- ☐ Provide daily check-ins and quick issue resolution
- ☐ Collect feedback on usability, speed, and quality
Phase 2: Early Adopters (Week 3-4)
Section titled “Phase 2: Early Adopters (Week 3-4)”- ☐ Expand to 25% of the team
- ☐ Create internal documentation (setup guide, FAQ, tips)
- ☐ Host a lunch-and-learn session
- ☐ Develop project-specific configuration (rules, CLAUDE.md, AGENTS.md)
- ☐ Set up a Slack/Teams channel for questions
Phase 3: Full Rollout (Week 5-8)
Section titled “Phase 3: Full Rollout (Week 5-8)”- ☐ Migrate remaining team members
- ☐ Provide formal training sessions
- ☐ Create video walkthroughs for common workflows
- ☐ Establish weekly office hours for the first month
- ☐ Monitor adoption metrics (usage, satisfaction, productivity)
Phase 4: Optimization (Month 3+)
Section titled “Phase 4: Optimization (Month 3+)”- ☐ Review usage patterns and adjust subscription tiers
- ☐ Share advanced techniques and best practices
- ☐ Develop team-specific custom commands and automation
- ☐ Measure ROI against pre-migration baseline
- ☐ Phase out old tool licenses
Configuration Migration
Section titled “Configuration Migration”Settings Transfer
Section titled “Settings Transfer”Cursor’s one-click import handles most settings. For manual transfer:
VS Code: ~/.config/Code/User/settings.jsonCursor: ~/.config/Cursor/User/settings.jsonCopy editor, workbench, terminal, and git settings. Add Cursor-specific settings after import.
Translate your .cursor/rules/ into CLAUDE.md:
## Project Overview[From your Cursor rules description]
## Coding Standards[From your Cursor rules patterns]
## Common Commands- Build: npm run build- Test: npm test- Lint: npm run lint
## Architecture Notes[From your Cursor rules architecture section]Create AGENTS.md from your existing configuration:
## Project[Description of the project]
## Standards[Coding conventions, patterns to follow]
## Commands- Test: npm test- Build: npm run build- Deploy: npm run deploy
## Rules[What to always do, what to never do]Common Challenges
Section titled “Common Challenges”Challenge 1: Muscle Memory
Section titled “Challenge 1: Muscle Memory”Problem: Reaching for old shortcuts and workflows out of habit.
Solution: Print the Keyboard Shortcuts reference. Keep it visible for 2 weeks. Use the command palette (Cmd/Ctrl+Shift+P) when you forget a shortcut.
Challenge 2: Trust Calibration
Section titled “Challenge 2: Trust Calibration”Problem: Not trusting AI suggestions (under-trusting) or accepting everything blindly (over-trusting).
Solution: Start with low-stakes tasks. Always review diffs before accepting. Use version control liberally. Gradually increase trust as you learn the tool’s strengths and weaknesses.
Challenge 3: Context Overload
Section titled “Challenge 3: Context Overload”Problem: Providing too much or too little context to the AI.
Solution: Start specific — include only the files directly relevant to the task. If the AI’s output is off, add more context incrementally. Use @ references and file scoping rather than “include everything.”
Challenge 4: Workflow Translation
Section titled “Challenge 4: Workflow Translation”Problem: Trying to use the new tool exactly like the old one.
Solution: Each tool has its own strengths. Do not force Claude Code to work like an IDE. Do not treat Cursor like a CLI. Embrace the tool’s native workflow — you will be faster once you adapt.
Success Metrics
Section titled “Success Metrics”Individual Metrics
Section titled “Individual Metrics”| Metric | How to Measure | Target |
|---|---|---|
| Time to feature | Hours per story point | -20% after 1 month |
| Bug rate | Bugs per release | -30% after 2 months |
| Code review time | Hours per PR | -40% after 1 month |
| Days to proficiency | Self-assessment + output quality | Under 14 days |
Team Metrics
Section titled “Team Metrics”| Metric | How to Measure | Target |
|---|---|---|
| Adoption rate | % of team actively using daily | >80% after 2 months |
| Satisfaction | Anonymous survey (1-5 scale) | >4.0 |
| Velocity change | Sprint velocity comparison | +25% after 3 months |
| Cost per developer | Tool subscriptions / developer count | Track monthly |
Rollback Plan
Section titled “Rollback Plan”Always have an exit strategy:
- Keep old tool licenses for 3 months after migration
- Export all configurations before making changes
- Define rollback triggers — measurable criteria that would prompt reverting
- Set evaluation checkpoints at 30, 60, and 90 days
- Keep the rollback decision objective — use metrics, not feelings
Final Checklist
Section titled “Final Checklist”- ☐ Pre-migration assessment complete
- ☐ Budget approved
- ☐ Pilot users identified and trained
- ☐ Internal documentation created
- ☐ Configuration files migrated (rules, CLAUDE.md, AGENTS.md)
- ☐ MCP servers installed and tested
- ☐ Team training sessions scheduled
- ☐ Success metrics defined and baseline measured
- ☐ Rollback plan documented
- ☐ Go-live date communicated
- ☐ Support channel established (Slack/Teams)
- ☐ Old tool licenses scheduled for cancellation (after 3-month overlap)
Frequently asked questions
How long does migrating to a new AI coding tool take?
Migration is a process that takes 2-3 months to complete fully, not a one-time event. The team plan runs a pilot in weeks 1-2, early adopters in weeks 3-4, full rollout in weeks 5-8, and optimization from month 3 onward.
What has to be decided before the migration starts?
The pre-migration assessment fixes which tools the team is moving from and to, how many developers are affected, the budget per developer per month, compliance requirements, the acceptable productivity dip, the pilot users, and what success looks like.
How do you move Cursor rules to Claude Code or Codex?
Translate .cursor/rules/ into CLAUDE.md, mapping the rules description to a project overview and the rules patterns to coding standards, then add common commands and architecture notes. The same content becomes AGENTS.md for Codex.
Which metrics show the migration worked?
Individual targets are -20% time to feature after 1 month, -30% bug rate after 2 months, -40% code review time after 1 month, and under 14 days to proficiency. Team targets are above 80% adoption after 2 months, satisfaction above 4.0, and +25% velocity after 3 months.
What belongs in the rollback plan?
Keep old tool licenses for 3 months after migration, export all configurations before making changes, define measurable rollback triggers, set evaluation checkpoints at 30, 60, and 90 days, and keep the decision objective by using metrics rather than feelings.