Claude Code

Anthropic’s deep-context AI coder

5.0•664 reviews•

11K followers

Anthropic’s deep-context AI coder

5.0•664 reviews•

11K followers

AI Coding Agents

•

Command line tools

•

AI Code Editors

Anthropic’s AI coding assistant, designed for deep context understanding and capable of handling complex software tasks with a massive context window (up to 200K tokens).

Claude Code Reviews

The community submitted 664 reviews to tell us what they like about Claude Code, what Claude Code can do better, and more.

5.0

Based on 664 reviews

Review Claude Code?

Reviewers broadly see Claude Code as a strong terminal-first coding agent that handles whole tasks, not just single files. They repeatedly praise its ability to keep large codebases and multi-step work in context, make cross-file changes, run tests, and fit into existing git-heavy workflows. Makers of Wispr Flow and Katalyst especially highlight repo-scale understanding and reusable team patterns. The main complaints are practical: session memory and compaction can lose context, usage and cost are hard to track, and it still needs careful planning, review, and skill to use well.

+673

Summarized with AI

Pros

Cons

Reviews

All Reviews

Most Informative

What's great

code generation (15)massive context window (5)complex software tasks (10)

Claude Code is an exceptional AI coding agent that excels across the full spectrum—from rapid startup SaaS builds to enterprise-grade, multi-layered, complex applications. When provided with proper context and guided by fundamental software architecture, engineering principles, and security standards, it consistently delivers high-quality results. Used with common sense and real development experience, there is currently no better AI coding agent in my opinion.

What needs improvement

Claude Code CLI is already seamless and consistently delivers high-quality results. The main area for improvement would be deeper scalability toward a full agentic development environment (ADE), similar to what tools like Warp are evolving toward—bringing more autonomous workflows, richer context management, and tighter developer-environment integration.

vs Alternatives

Codex by OpenAI

Grok Code Fast 1

I evaluated Warp, OpenAI Codex, and Grok Code Fast1, but Claude Code stood out for its balance of control, context awareness, and consistent output quality. It scales equally well from rapid prototyping to complex, enterprise-grade systems, while remaining predictable and effective when guided by solid engineering and security practices—making it the most reliable choice overall.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

553 views8mo ago

What's great

code generation (15)massive context window (5)complex software tasks (10)ai agents (2)

I've built multiple enterprise apps with Claude Code. Not prototypes — actual production systems with payments, auth, real-time features, the lot. I'm building all day every day and it genuinely keeps up. Most AI coding tools feel like autocomplete with extra steps. Claude Code feels like having a senior dev sitting next to you who actually understands context. It reads your codebase, remembers your patterns, and suggests things that make sense for YOUR project, not generic boilerplate.

What needs improvement

memory limitations (1)

You need to learn advanced practices so that you can make the most out of this tool if you don't you won't multiply your productivity.

vs Alternatives

Copilot is good for single-line completions but falls apart on anything complex. Cursor is decent but I kept hitting walls with context it'd lose track of what I was building. Claude Code just gets it. I can describe a feature in plain English, point it at the right files, and it produces code that actually works within my existing architecture. The difference is night and day once your project gets past a few hundred lines.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

432 views6mo ago

What's great

code generation (15)massive context window (5)

Honestly, it just works. You point it at a real codebase, give it a messy multi-step task, and it actually finishes: reads the right files, edits them, runs the tests, reports back. The skills system is the part I didn't know I needed: instead of hoping the model behaves well, you literally wire in the discipline (TDD, brainstorming, code review) and it just follows it. Plan mode + sub-agents mean I can scope big work upfront and let it run in the background while I do something else. And the memory thing? Game changer. I tell it once "don't use uv in this repo" and it remembers next time without me repeating myself.

What needs improvement

high pricing (1)

The learning curve is real. Handing a new dev the docs and saying "good luck" is a bit mean. It can also be expensive if you let sub-agents rip; I wish the spend was more visible in real time. Sometimes it re-reads files it just edited or runs a grep when it could've used a better tool, small papercuts. The permission prompts get old fast on safe read-only stuff. And when a hook or skill silently misbehaves, debugging it is a headache, give me a verbose trace mode, please. Oh, and localize the system messages properly; the model speaks Russian fine, the UI doesn't.

vs Alternatives

Codex 3.0 by OpenAI

I tried most of the alternatives. Cursor is great if you live in the IDE, but I'm a terminal person. Copilot is fine for autocomplete but feels shallow when you ask it to do something real. Aider is lovely and open-source, but I kept running into its limits on bigger reasoning tasks. Cody is solid for searching a giant repo, weak for actually changing things. The direct API is the most flexible and the most painful, because you build all the glue yourself. Gemini/ChatGPT in the browser are nice for thinking out loud, useless for touching code. Claude Code hit the sweet spot for me: smart model, real tools, and a harness that doesn't get in the way.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

148 views3mo ago

What's great

Most of my building happens here now rather than in an editor, which is not a sentence I expected to write.

The thing that actually changed my workflow was not code generation. It was that the unit of work got bigger. I stopped asking for functions and started handing over outcomes, and the useful skill turned out to be decomposition rather than prompting. Say what done looks like, say what must not break, then get out of the way.

Two things I would tell someone starting:

Ask it how it would do the thing before you ask it to do the thing. The plan is cheap to read and expensive to skip. Most of my bad runs were ones where I skipped straight to execution.

Give it a memory. Out of the box every session starts from zero and you re-explain the same context forever. A docs folder it reads on every run was the single biggest improvement I made, and it cost nothing.

Where it bites: long sessions quietly compact and you lose reasoning you did not know you needed. And it is confidently wrong in exactly the register it is confidently right, so the review burden never really goes away, it just moves.

Still, it replaced the thing I used to call my job.

What needs improvement

Context that does not survive a session is the big one. Every new run starts from zero on decisions I already made, and the workaround is a memory folder I maintain by hand. That should not be my job.

Related: auto-compaction happens quietly. I lose reasoning without being told which reasoning went. Even a line saying what got dropped would let me decide whether to care.

Cost visibility is the third. I route deliberately, cheap models to gather and expensive ones to judge, and I have no way to see whether that routing pays for itself. The spend is real and the feedback loop is invisible.

Smallest one: it is equally confident when it is wrong. Not a solvable problem exactly, but a calibration signal on the output would change how much I re-read.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

71 views1mo ago

What's great

massive context window (5)complex software tasks (10)

It's the only one of these that holds a whole task instead of a whole file. I hand it a failing test and a rough intent, and it goes and reads the three files it actually needs, which is the part I used to do by hand before any code got written. Terminal-first matters more than it sounds too, because the thing lives next to my git history and my scripts rather than inside an editor's idea of what my project is.

What needs improvement

context compact (1)

The usage limits are the weak spot. A percentage tells me nothing on its own because the limits are rolling windows, and the decision I'm actually making at 78% is start the big refactor or go make coffee. Only the reset clock answers that, and it isn't in front of me.

vs Alternatives

I run more than one on purpose, Codex and Cursor both get used here. Claude Code wins when the plan matters more than the completion, anything touching four files and a migration. For a single file the others are quicker and I still reach for them there.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

46 views1mo ago

What's great

code generation (15)

Thanks to the Claude Code team for building such a great product — it really makes a software engineer’s life easier. It is impressive. Once you clearly define the problem, it often delivers an almost perfect solution — sometimes even better — especially if you already have a few unit or integration tests in place. In most cases, with just a bit of debugging and some error context, it gets about an 85% approval rate from senior engineers.

What needs improvement

It still struggles a bit with frontend web code, but that’s mostly because frontend details are harder to describe precisely and harder to verify automatically.

vs Alternatives

almost perfect if you're clear about the solution. almost hand free once you describe the requirement clear. comparing to other , approve rate is much higher than others.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

280 views10mo ago

What's great

code generation (15)

Claude is excellent for writing clean code and reasoning through complex logic. It’s especially good at understanding context and improving code quality.

What needs improvement

Better support for very technical edge cases and framework-specific examples would be helpful.

vs Alternatives

I considered other AI coding assistants, but Claude felt more reliable for structured reasoning and reviewing larger code blocks.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

127 views7mo ago

What's great

I don't normally leave a review but Claude Code is by far the best use of any of the LLMs so far. It's really capable and is always improving. Tested with html, css, javascript, Xcode, Android Studio, command line, Claude Desktop, Visual Studio Code and it is really a seamless experience and the quality of the good is good.

What needs improvement

Handling of large code databases especially after compacting could be better.

vs Alternatives

ChatGPT by OpenAI GPT-5

Tested the same prompts for a simple html website and the comparison was laughable. Claude Code literally created what I had in my head with visuals and styling that worked well whilst Chat GPT just produced some text and a hyperlink for the page.

Ratings

Ease of use

Reliability

Value for money

Customization

Report

73 views5mo ago

What's great

code generation (15)

I was sceptic to work in my terminal with AI, but now, oh my. I've tried a lot of ai tools but Claude does what I need.

What needs improvement

Well, we all want it as cheap as possible?

vs Alternatives

Codex, OpenAI and lots of others. But Claude code fit my needs and style

Ratings

Ease of use

Reliability

Value for money

Customization

Report

5 views1mo ago