CodiumAI Ltd., a startup that has created a generative artificial intelligence-powered code testing tool for developers, said today it’s making life easier for its users with the launch of a new “proactive AI Agent.”

The new tool, somewhat similar to GitHub Inc.’s Copilot tool, can instantly generate code to fix any errors or inefficient code it comes across.

The Israeli startup first hit the headlines exactly one year ago when it launched its platform, backed by an $11 million seed funding round. It’s main offering is the CodiumAI Coding-Agent, which is powered by OpenAI’s GPT-4. The idea is that Coding-Agent helps developers write more efficient code by testing it as they write it.

CodiumAI reckons it’s solving a big problem in the software industry, citing a 2020 study that shows how bad quality software costs U.S. companies more than $2 trillion annually.

Code testing is a key part of the software development process, and the startup enhances this aspect by enabling developers to take a more proactive approach. Traditionally, code is only tested for quality after it has been written, as it’s a laborious and time-consuming task. Moreover, it’s an imperfect process, as developers don’t always know what they should be testing their code for.

Coding-Agent works by analyzing newly written code as the developer is working. It suggests which tests should be performed to ensure the code functions as it should. All the developer has to do is accept and commit those tests, and Coding-Agent will quickly inform them if it’s up to scratch.

Now, CodiumAI is going one step further. With the launch of its AI Agent, it doesn’t just suggest and run code tests, but can also provide recommendations to fix any bad quality code that it discovers. The AI Agent works in tandem with popular code completion tools such as GitHub’s Copilot. The startup’s own code completion tool, launched today, places more emphasis on code integrity.

It says its new offering is highly relevant because more than 92% of developers are now using generative AI tools to assist them in writing code. But although these tools can help humans to write code faster, there are a lot of concerns about the quality of AI-generated code. This is what CodiumAI intends to fix.

The startup said its new agent runs alongside developers as they work, suggesting relevant code tests, drafting documentation for their code and identifying any duplicates without being prompted, as well as providing suggestions to tidy up any bad code it finds.

CodiumAI co-founder and Chief Executive Itamar Friedman said developers are looking for AI tools that can assist them with the tedious tasks that are essential for ensuring high quality code and correctness. “We are looking to an ‘agentic future’ where busy developers will team up with agents to improve code quality and processes,” he said. “Integrated agents will be there every step of the way, offering insights, recommendations and solutions proactively.”

Image: CodiumAI