AI

In a sea of agents, AWS bets on structured adherence and spec fidelity

Despite new methods In these emerging developments, companies continue to turn to autonomous coding agents and code generation platforms. Competition to get developers to work on their platforms, coming from tech companies, has also increased.

AWS believes its offering, Kiroand new capabilities to ensure behavioral fidelity provided a major differentiator in the increasingly crowded coding agent space.

Kiro, first launched in July on public preview, is now generally available with new features including property-based testing for behavior and a command-line interface (CLI) to customize custom agents. Kiro is an agentic coding tool with its own IDE that allows creating agents and applications from prototype to production.

Deepak Singh, AWS vice president of developer agents and experiences, told VentureBeat that Kiro “keeps the fun” of coding while providing structure.

“As I like to say, what Kiro does is it allows you to talk to your agent and work with your agent to build software just like you would with any other agent,” Singh said. “But what Kiro does is it takes this structured way of writing that software, which we call spec-driven development, into specifications that turn your ideas into things that last over time. So the result is more robust, maintainable code.”

In addition to new features, AWS is offering startups in most countries one year of free credits for Kiro Pro+ and expanded access to Teams.

Behavioral therapy and checkpoints are built in

One of Kiro’s new features is property-based testing and checkpointing.

One problem some companies face with AI-generated code is that it can sometimes be difficult to assess accuracy and how well the agents are adhering to their intended purpose. AWS noted in a blog post that “whoever writes the tests (human or AI) is limited by their own biases – they have to think of all the different, specific scenarios to test the code against, and they will miss edge cases they hadn’t thought of. AI models often ‘game’ the solution by tweaking tests rather than fixing code.”

See also  Mistral AI’s new coding assistant takes direct aim at GitHub Copilot

“What feature-based testing does is it takes a specification, it takes a specification, and based on that it identifies the properties that your code should have, and it actually creates potentially hundreds of test scenarios to verify that your code does what you intended as identified in the specification, and it does all that automatically,” Singh said.

Singh said organizations can upload their specifications and the Kiro agent can start identifying what’s missing even before the code review process begins.

Property-based testing links the specified behavior, also called your instructions, to what the code does. Kiro can help users write it into their specifications based on EARS format. For example, if a company is building a car sales app, the specification would be:

“For each user and each car list, WHEN the user adds the car to favorites, THE System MUST display that car in its favorites list. PBT then automatically tests this with User A adding car #1, User B adding car #500, User C adding multiple cars, users with special characters in user names, cars with different statuses (new, used, certified), and hundreds of other combinations, noting edge cases and verifying that the implementation matches your intentions.”

Unlike a traditional unit test specification, which states: if a user adds car #5 to their favorites, it will appear on their list.

Kiro will then identify examples of the code that violates the specifications and present them to the user.

Kiro now also enables checkpointing, so developers can roll back to a previous change if something goes wrong.

CLI coding

The second major new feature of Kiro is Kiro CLI, which puts the Kiro coding agent directly into a developer’s CLI.

See also  How much information do LLMs really memorize? Now we know, thanks to Meta, Google, Nvidia and Cornell

AWS said the Kiro CLI uses some features of the Q Developer CLI: the in-line coding assistant, launched in October 2024—to allow users to access the agent from the command line.

It also allows developers to build custom agents, such as a backend specialist, a frontend agent, and a DevOps agent, tailored to an organization’s codebase.

Singh said developers have their own unique ways of working, so it’s important that coding agent providers like AWS meet them wherever they are. Kiro CLI allows users to:

  • Stay in the terminal without having to switch contexts

  • Structuring AI workflows with custom agents

  • Have one set up for two environments, as MCP servers and other tools work in both the Kiro version on the IDE and in the CLI

  • Fast automation to format code or manage logs via automated commands

Coders competition

However, Kiro is just one of many coding platforms popping up and competing for business use.

By OpenAI’s GPT-Codex, which unifies its Codex Coding Assistant with IDEs, CLIs and other workflows, to Googling’With the Gemini CLI, it’s clear that more and more developers are demanding easy access to coding agents where they do their work.

And companies are placing increasing demands on coding agents. For example, Anthropic made his Claude Code platform available on the internet and mobile. Some coding platforms also allow users to choose which model they want to use for their coding.

Singh said Kiro does not depend on just one LLM; instead, it guides to the best model for the job, including AWS models. At launch in July, Kiro was based on Claude Sonnet 3.7 and 4.0. The current iteration uses Claude Sonnet 4.5 and Haiku 4.5. Well-known brands such as monday.com to have noted the significant benefits of AI-powered coding, showing that companies are likely to continue using these platforms in the future.

See also  Expired listings: an agent's goldmine

“We saw that the mental model for developers is changing, but it’s not just about becoming more efficient; it’s also about the way they organize around the way they work now,” says Singh.

Source link

Back to top button