Last week, Anthropic dropped a preview of new functionality it calls Claude Code — which the company describes as “a command line tool for agentic coding.” It enables a bunch of useful functions for developers including searching and reading code, editing files, testing, committing and pushing code to GitHub and so on. This was delivered in conjunction with the Claude Sonnet 3.7 model release. The combined timing is very smart because Sonnet is a favorite model among application developers.
Why Sonnet Is A Developer Favorite
Sonnet’s popularity comes from its ability to “reason” — meaning that it’s allowed to spend more time examining various datapoints before answering — paired with its model training specifically for development tasks. This value proposition seems to be working, given that Sonnet is one of the more expensive models out there, yet developers are willing to pay the premium — in many cases out of their own pockets.
The 3.7 version is interesting in that it takes a new hybrid approach to reasoning, in which responses are based upon the nature of the question. This is especially useful to a developer who may at times ask to do something simple (like debug a line of code) but at other times want to achieve something complex (like building a new app from scratch). Anthropic also claims that this new version is the top performer in developer-specific AI benchmarks. All of these factors have led to quick adoption of Sonnet 3.7 in AI developer assist technologies from GitHub, AWS, Eclipse and others. So, Anthropic’s decision to preview this new developer tool alongside the new Sonnet model seems to be a smart move.
Do We Really Need Another IDE?
Given that Sonnet is a popular model and that it’s already exposed in the most popular toolsets, introducing a new IDE — which is effectively what Claude Code is — may seem curious. Since Anthropic monetizes Sonnet based upon millions of tokens served, it does not really need an IDE to be commercially viable. This raises the question: What does Claude Code have to offer that other IDEs paired with Sonnet don’t? Three answers jump out at me.
It’s lightweight. First, Claude Code is a very lightweight tool. It’s invoked by a command-line interface or an API. This aligns with more advanced developers seeking something very simple to use. The API is also nice for enabling devs to automate repetitive or periodic tasks without the use of any toolset.
It’s conversational. Like other AI assistants, Claude Code can be pointed at a codebase and provide documentation and insights to a new developer. Then the developer can ask Claude about the app being coded through either very specific or broader questions, thanks to Sonnet 3.7. But, crucially, this can happen without the distraction of tools and widgets and everything else in between. In some ways, this enables a more detached method for developers to understand the code and its possibilities before diving in. But the conversational approach also opens up code-related tasks such as updates to non-developers. This could include support staff or members of the DevOps team who don’t necessarily want to get into the weeds but who do need to execute certain tasks.
It’s cloud-agnostic. Claude Code is running in a terminal window and using your local codebase that you have pulled from source. So you don’t have to set up a lot of configuration parameters with other cloud providers or use their specific libraries and tools.
Claude Code Is About More Than Features
While the addition of generative AI assistants is providing some incremental productivity gains, the base IDE concept and the corresponding development process remain unchanged. But Claude Code is an attempt to rethink the IDE from an AI- or agentic-first perspective. This is what makes Claude Code important, because it’s very different from what we have seen so far. Keep in mind that it’s still in preview and may never get to production status; that said, it does suggest new possibilities.
After digging into this topic, I started to consider the following questions about the future of developer tools.
- Will LLM providers go “platform”? At this point, one has to wonder where the commercial LLMs will go next. In the span of about two years, we have seen rapid commoditization in the LLM market and also some possible limits to LLMs’ pace of innovation. So where can highly funded companies like Anthropic invest to grow? Claude Code may be a clue. Instead of being content to remain an ingredient in other companies’ platforms, we may see AI companies of this caliber become next-gen platforms on their own.
- What’s the future of low-code and no-code in the AI world? The addition of AI into the lives of developers has been profound. But we are also seeing AI provide significant augmentation to low-code and no-code platforms. In fact, the addition of AI has redefined the space, to the point that specialized user-experience or agentic capabilities like those embedded in ServiceNow or Zoho can be considered no-code platforms. While Claude Code is not low-code per se, it’s similar in that it does lower the learning curve of coding. Given that we are seeing this in so many enterprise and consumer product categories, it raises the question of whether low-code and no-code are still a distinct product category, or just table-stakes features that we can expect to see across many product categories.
- Is the future of dev tooling consolidation or fragmentation? Over the years there has been this vendor notion of creating a “sacred toolchain” or even just a standard toolset for developers. But the reality is far from that. Developers tend to use many tools and utilities to get their jobs done. Some of this is driven by pure necessity, such as the need to accommodate a legacy application with a proprietary toolset. Some of it is simply preference or past experience. And in the age of agentic AI, one must wonder if the notion of consolidated toolchains, tools and plugins is going to give way to a set of collaborative agents. These agents would effectively abstract the code and work in concert with developers and DevOps resources. And maybe Anthropic is seeing it that way, too, when we notice that Claude Code is categorized under “Agents and Tools” in the Anthropic documentation. The disruptive potential of an agentic IDE is an analog to Satya Nadella’s bold prediction that SaaS will collapse in the Agentic era.
To close, I’ll note that, while there are some novel capabilities that come with Claude Code, it’s not (yet) a threat to the other IDEs. And sure, it’s probably worth trying it out for some utility-based tasks and maybe even subscribing if and when it becomes generally available. But even if it doesn’t take off as a product, Claude Code’s real impact may be as a sign of things to come — specifically, how developers will work in the coming years.
Moor Insights & Strategy provides or has provided paid services to technology companies, like all tech industry research and analyst firms. These services include research, analysis, advising, consulting, benchmarking, acquisition matchmaking and video and speaking sponsorships. Of the companies mentioned in this article, Moor Insights & Strategy currently has (or has had) a paid business relationship with AWS, Microsoft, ServiceNow and Zoho.
