Claude Code Reviews & Product Details

What I like best about Claude Code is its ability to understand a codebase and help with larger development tasks rather than just generating isolated code snippets. It is useful for debugging, refactoring, explaining existing code, and implementing features across multiple files. It can save significant time when working on complex React Native projects. Review collected by and hosted on G2.com.

What I dislike about Claude Code is that it can sometimes make changes that need careful review, especially when working across multiple files. It can also misunderstand project-specific requirements or make assumptions about the existing codebase. I usually review and test its changes before committing them. Review collected by and hosted on G2.com.

I like Claude Code’s natural voice. It sounds less robotic and less compressed out of the box than some other LLMs. I really like its technical reasoning, especially how well it handles complex logic and refactoring. I also like how it handles project artifacts, with side-by-side comparisons and document rendering that can use workspace knowledge from our organization’s files. It plays well with the tools our agency uses and gives us a one-stop-shop workflow: it can generate and deploy code, help with infrastructure setup, and handle a lot of the work around our projects in one place instead of having us jump between apps and logins. Setup was very easy and relatively simple, though it did require some technical tooling and CI/CD pipeline setup. Review collected by and hosted on G2.com.

I did need to set up some technical tooling and our deployment CI/CD pipeline, though the setup was very easy and otherwise relatively simple. Review collected by and hosted on G2.com.

I like that Claude Code uses agentic AI to automate tasks I usually do manually in my daily engineering workflow. Its CLI harness is reliable and bankable, and I like that it’s backed by solid frontier models from Anthropic, such as Claude Sonnet and Claude Opus. I’m comfortable using CLI tools, and Claude Code’s CLI-based approach works well for me. It’s also feature-rich, with granular ways to customize and optimize my workflows. Setup was very smooth, and I didn’t run into any challenges installing it on our computers. Review collected by and hosted on G2.com.

I sometimes see flakiness with Claude Code staying authenticated. It gives me a 401 error and keeps retrying; after a bunch of retries, it starts working again. I think it needs a way to capture this in the audit trail, find the root cause of these intermittent authentication issues, and address the problem in the code. Review collected by and hosted on G2.com.

Claude Code is really fast, and its agent use is very optimized. I can run many agents at once, which boosts my efficiency while coding. I like the Opus model because it’s very smart and good at coding, and the UI makes it easy to interact with agents. It’s easy to develop and integrate with my other products. Environment management is good, and I like that agents can’t do anything without my permission. The MCP connections and connectors are implemented really well, so it’s easy to connect tools like Linear and Slack and use them with my agents. Setup was very easy and only took a few minutes. Review collected by and hosted on G2.com.

Sometimes I can't see changes in worktrees very well. When I click the worktree changes, it shows other changes that aren't relevant to what I'm working on, and sometimes it mixes them up with my local computer changes. I think it's a bug, and I'd like to see only the changes specific to that worktree. Review collected by and hosted on G2.com.

It treats me like a developer without requiring me to be one. I'm a photographer and content strategist by trade. I don't have an engineering background. But over the past year I've used Claude Code to build a proprietary WordPress plugin stack, a custom SEO diagnostic engine that connects to five different data APIs, and an entire custom directory platform for a tourism website, all of which are live and serving real users right now. The thing that separates Claude Code from every other AI coding tool I've tried is that it understands the full context of what you're building, not just the file you're looking at. It holds the architecture in its head, catches conflicts between components before they break, and writes code that actually works on the first or second pass more often than it has any right to. I spent two years trying to make off-the-shelf directory plugins work for my business and failing. Claude Code helped me build something better from scratch in a fraction of that time. Review collected by and hosted on G2.com.

Two things, both manageable but worth knowing about.

First, drift. On longer or more complex projects, Claude Code can lose the thread of what you've already established and start making decisions that contradict earlier work. When you're building a plugin stack with multiple interconnected components, that means you sometimes have to re-explain context that should already be understood. It's not a dealbreaker, but it means you need to be assertive about holding the line on your architecture. The tool works best when you treat it like a very talented junior developer who occasionally needs to be reminded what was already decided.

Second, there's a tendency to reach for patches instead of solutions. When something breaks or a feature doesn't work as expected, the first instinct is sometimes to bandage the immediate symptom rather than address the underlying structure. I've learned to be specific about wanting root-cause fixes, not workarounds. Once you communicate that expectation clearly, the output quality goes up significantly. But it took some reps to figure out that I needed to ask for it that way.

Neither of these has stopped me from shipping real production software with the tool. They're just the learning curve of working with it effectively. Review collected by and hosted on G2.com.

The slash commands. They leave me with the ability to combine the solutions to manual coding that claude code already combines, with a solution to the biggest problem I have with chatbots: conversation flow and context. What I mean by this, is that when you ask a question, the ai will take your string of words and run with it. In claude codes case, / commands let you specify your intent in the query and it really should be implemented into regular chats, such as /btw and /plan. Makes the workflow closer to a regular conversation Review collected by and hosted on G2.com.

Nothing. If I had to say one thing I suppose it would be understanding of human unrefined thinking. If you have a rigorously structured and specified prompt, claude code can complete almost any task it is able to touch; however, LLMs are a complicated machine and isn’t able to “read my mind,” so to speak. I don’t expect it to, as in this situation I cannot fully express the extent of my thoughts. It Claude code could add some kind of form before building new products as a feature for beginners, I think it would really help bring clarity and individuality to vibe coders Review collected by and hosted on G2.com.

I use Claude Code to automate form filling, search emails, gather documentation, and speed up manual tasks my team members can do with Python code. It reduces the time I spend on manual data entry. It’s very quick, troubleshoots and debugs on its own, and can read emails and analyze documents. The initial setup was easy. Review collected by and hosted on G2.com.

There is nothing that I dislike. Review collected by and hosted on G2.com.

The harness, models, plugins, skills, the intelligence, speed, guardrails, updates Review collected by and hosted on G2.com.

Occasional connectivity hiccups—it keeps throwing “can’t connect to APIs” errors. This happens most often when I’m moving between places. I really wish it were smart enough to pause the work when the network gets choppy or unavailable, and then automatically resume once connectivity is back. Review collected by and hosted on G2.com.

Lives in your terminal, which is great if you live there too - pity the best commands are buried where new users won't find them.

Plays beautifully with MCP, Skills, and your existing connector stack, though every MCP call quietly piles tokens onto the bill.

Genuinely strong on big refactors and the 1M context is no gimmick. Just pin your version, because v2.1.100+ has been burning tokens faster than it should.

Pro at $20 looks tempting but you'll blow through it in one serious session; Max or Team Premium is where the maths actually works.

Docs are good and the community fills the gaps, but onboarding is very much figure-it-out-yourself, and Anthropic's occasional tier experiments don't help procurement conversations.

Closest thing yet to working with an actual junior engineer rather than glorified autocomplete. Review collected by and hosted on G2.com.

You get addicted to using it, and then you realize you can’t work without it. Review collected by and hosted on G2.com.

What I like most about Claude Code is how good it is at understanding larger context and structured thinking during development. It is especially useful for reviewing codebases, reasoning through architecture decisions, refactoring, and generating cleaner code with less back and forth. The responses also feel more natural and less robotic compared to many coding assistants. Review collected by and hosted on G2.com.

One downside is how quickly it consumes tokens, especially on larger projects. It also tends to reread and rebuild context from scratch quite often, which can feel inefficient during long development sessions. While the reasoning quality is strong, context management and continuity could definitely be improved. Review collected by and hosted on G2.com.