Step 1: Explore
Reads files, searches patterns, and builds a mental model of your project structure and dependencies
Step 2: Plan
Designs an implementation strategy, identifies files to change, and considers edge cases
Step 3: Code
Edits existing files, creates new ones, updates imports, and runs commands across your project
Step 4: Verify
Runs tests, checks for build errors, and validates the result. Loops back if anything fails
0x
More deployments (Vercel)
0%
Faster investigations (Ramp)
0x
Revenue growth (Vercel)
What do these enterprise numbers mean?
These statistics come from public statements by Vercel and Ramp about their real production usage of Claude Code. Vercel, the company behind Next.js, reported that their team deployed 7.6 times more frequently after adopting Claude Code. Ramp, a corporate card and expense management company, reported that error investigations took 80% less time. These are production metrics from engineering teams, not synthetic benchmarks.
| Tool | What it does | |
|---|---|---|
| Read | Read | Read any file in your project: code, config, docs, even images |
| Edit | Edit | Make precise string replacements in existing files |
| Write | Write | Create entirely new files when needed |
| Bash | Bash | Run shell commands: build, test, git, npm, docker, anything |
| Glob | Glob | Find files by name pattern (e.g., **/*.tsx) |
| Grep | Grep | Search file contents with regular expressions |
| Task | Task | Delegate work to subagents that run in parallel |
How to get the best results
Claude Code works best when it has context. Start by running
/init to generate a CLAUDE.md that describes your project. Be specific in your requests: "Add a dark mode toggle to the settings page that persists to localStorage" gives much better results than "add dark mode." For complex tasks, use Plan Mode (Shift+Tab) to review the approach before Claude Code writes any code.