treeru.com
Tools

Claude Code Hands-On Review: AI That Writes Your Code From the Terminal

AI coding tools are everywhere, but how many developers have actually used one on real projects for months? I spent 3 months using Claude Code in production — writing 23 blog posts, optimizing PageSpeed scores, and automating server management. Here is the honest review.

23

Blog Posts Written

3 mo

Usage Period

CLI

Interface

~70%

Productivity Gain

1What Is Claude Code

Claude Code is a terminal-based AI coding assistant built by Anthropic. Unlike VS Code extensions or IDE plugins, it runs directly in the terminal. Point it at your project folder, type a natural-language instruction, and it reads files, edits code, creates new files, and even runs git commits.

Core Capabilities

Natural-language coding instructions in the terminal
Full project context awareness with multi-file editing
Direct integration with git, npm, docker, and other CLI tools
Create, modify, and delete files autonomously
Code review, refactoring, and test generation

Why CLI Instead of an Editor?

GUI editors are optimized for working on one file at a time. A CLI tool can operate on the entire project at once. That is why requests like "create 5 blog posts in the existing pattern" actually work.

2CLI-Based Workflow

After installation, you can use it immediately from the project root. No config files needed — it automatically analyzes the project structure.

# Basic usage

# Start interactive mode in the project folder
claude

# Give a direct instruction
claude "Add a dark mode toggle to the Header component"

# Target a specific file
claude "Change the session expiry in lib/auth.ts to 24 hours"
CommandDescriptionUse Case
claudeEnter interactive modeComplex multi-turn tasks
claude "instruction"Single-shot executionQuick fixes, simple additions
/helpView built-in helpSee available commands
/commitCommit changes (slash command)Auto-commit after work
/review-prReview a PR (slash command)Automated code review

Tip: Specificity matters

"Fix the code" gives vague results. "In Header.tsx, change the active nav item style to use the primary color" gives precise edits. Always include file names, locations, and expected outcomes for best results.

3Real-World Use Cases

I used Claude Code across three primary domains over the 3-month period.

Blog Content

23 posts written (TSX)
Auto-generated metadata
SEO-optimized structure
Batch series updates

Performance

PageSpeed score 38 to 88
Image compression automation
Bundle size optimization
Layout shift elimination

Server Ops

Monitoring scripts
Cron job configuration
Log analysis automation
Deploy script improvements

# Actual prompt to generate 5 blog posts at once

claude "Create 5 blog posts.
Each post has _meta.json, page.tsx, layout.tsx, and 1s.tsx.
Category is dev, with unique tags and related posts for each.
Follow the same pattern as existing blog posts."

Productivity Impact

Writing a blog post manually took 2-3 hours. With Claude Code, that dropped to under 30 minutes. Review and edits are still needed, but first-draft speed is incomparable.

4Key Strengths

After 3 months of daily usage, these strengths stood out. Claude Code particularly shines for large-scale operations.

StrengthDescriptionPractical Impact
Multi-file editingModifies multiple files in a single pass80% reduction in refactoring time
Project awarenessUnderstands the entire codebaseConsistent code style across files
Batch automationProcesses patterned tasks in bulk20+ blog posts generated rapidly
CLI tool integrationRuns git, npm, docker directlyZero workflow interruption
Context retentionRemembers prior work within a sessionComplex multi-step tasks handled

Top Strength: Large-Scale Repetitive Tasks

When you need to create 10 or 20 files following the same pattern, Claude Code truly excels. Humans inevitably make mistakes in repetitive work, but the tool delivers consistent quality every time. For template-based code generation, it is effectively more accurate than a human.

5Limitations and Caveats

Claude Code is not a silver bullet. There are clear gaps and areas requiring caution.

Write security-critical code yourself

Authentication, payment processing, and PII handling logic should never be copy-pasted from AI output. Security vulnerabilities and missed edge cases are a real risk.

Cost management is essential

Token-based pricing means large projects can accumulate costs quickly. Define scope clearly before starting, and minimize unnecessary context.

Code review is mandatory

Never trust generated code blindly. Especially with complex business logic, code that appears to work correctly may contain subtle logical errors.

Limitation AreaSpecific ExampleWorkaround
Complex business logicPayment flows, state machinesDesign first, use AI for implementation only
Design senseVisual layout, UX decisionsImplement from design mockups
Latest librariesNew API breaking changesCross-reference official docs
Large-scale refactoringFull architecture overhaulsBreak into smaller incremental steps

6Claude Code vs Cursor

The most common question I get is "How is it different from Cursor?" Having used both extensively, here is the breakdown.

FeatureClaude CodeCursor
InterfaceCLI (Terminal)GUI (VS Code fork)
Multi-file editingVery strongModerate
Learning curveCLI experience neededLow
Repetitive tasksExcellentModerate
Real-time autocompleteNoneYes (Tab)
Server workSSH-capableLocal-focused
PricingUsage-basedMonthly subscription
Best forCLI-proficient devsEditor-centric devs

Claude Code Wins When

Editing 10+ files simultaneously
Working directly on servers via SSH
Bulk-generating patterned content
CI/CD pipeline tasks

Cursor Wins When

Real-time code autocomplete
Quick single-file edits
GUI-based comfortable workflow
AI coding beginners

Verdict: Use Both

Cursor for single-file work, Claude Code for project-wide operations. Switching tools based on the task is the realistic optimal strategy. You do not have to pick just one.

Key Takeaways

  • Claude Code is a terminal-based AI coding CLI, strongest in multi-file operations
  • Validated across 23 blog posts, PageSpeed optimization (38 to 88), and server management
  • ~70% productivity improvement in bulk pattern generation
  • Security-critical code and complex business logic still need manual writing
  • Claude Code and Cursor are complementary — switch tools by task type

This article is based on Claude Code usage experience as of February 2026. AI tool features and pricing are subject to change. Actual results may vary depending on project characteristics. Always have security-critical code reviewed by qualified engineers. Non-commercial sharing of this content is welcome. For commercial use, please reach out via our contact page.