Amazon Web Services Introduced on Wednesday Caro Powersa system that allows software developers to give their AI coding assistants quick, specialized skills in specific tools and workflows — identifying what the company calls a major disruption in how artificial intelligence agents work today.
AWS made the announcement at its annual Answer: Invention Conference In Las Vegas. This capability marks a departure from how most AI coding tools work today. Typically, these tools keep every possible feature in front of memory — a process that burns through computational resources and can overwhelm the AI with irrelevant information. Cero Powers takes the opposite approach, activating specialized knowledge only when a developer actually needs it.
"Our goal is to give the agent specialized context so that it can reach the right results faster – and in a way that minimizes costs, too." Deepak Singh, vice president of developer agents and experiences at Amazon, said in an exclusive interview with VentureBeat.
The launch includes partnerships with nine technology companies: Datadogfor , for , for , . Dytracefor , for , for , . Figmafor , for , for , . Neonfor , for , for , . Net Leaffor , for , for , . The postmanfor , for , for , . stripfor , for , for , . Soup baseand AWS’s own services. Developers can also create and share their options with the community.
Why AI Coding Assistants Choke When Developers Combine Too Many Tools
To understand why Caro Powers matters, this AI development tool helps to understand the growing tension in the market.
Modern AI coding assistants rely on something called Model context protocolor MCP, to connect to external tools and services. When a developer wants their AI assistant to work with Stripe for payments, Figma for design, and Soupbase for databases, they connect MCP servers for each service.
Problem: Each connection loads dozens of tool definitions into the AI’s working memory before writing a single line of code. According to the AWS documentation, connecting just five MCP servers can consume more than 50,000 tokens—about 40 percent of the AI model’s context window—the developer also typed in his first request.
Developers are increasingly vocal about the issue. Many have complained that they don’t want to burn through their token allocation just to have an AI agent figure out which tools are relevant to a particular task. They want to jump right into their workflow — not watch an overloaded agent struggle to sort out irrelevant context.
This trend, which some in the industry call "context," This leads to slower response times, lower quality results and significantly higher costs—since AI services are typically charged via tokens.
Inside technology that loads AI expertise on demand
Caro Powers addresses this by packaging three components into a single, dynamically loaded bundle.
The first component is a steering file called power.md, which serves as an onboarding manual for the AI agent. It tells the agent what tools are available and, importantly, when to use them. The second component is the MCP server configuration itself – the actual connection to external services. The third includes optional hooks and automations that trigger specific actions.
When a developer is mentioned "payment" or "Check out" In his conversation with Caro, the system automatically activates the power of the strip, and loads its tools and best practices into context. When the developer shifts to database work, Supbase is enabled while Strip is disabled. Using the baseline context Zero when no options are enabled.
"You click a button and it loads automatically," Singh said. "Once a build is powered up, developers simply select ‘Open in Keru’ and it launches the IDE with everything ready to build."
How AWS is bringing elite developer techniques to the masses
Singh developed Kero Powers as a democratization of modern development practices. Before this capability, only the most sophisticated developers knew how to properly configure their AI agents with specialized contexts—writing custom steering files, generating precise signals, and manually managing which tools were active at any given time.
"We have noticed that our developers are adding capabilities to make their agents more proficient," Singh said. "They wanted to give the agent some special powers to do a particular problem. For example, they wanted their front-end developer, and they wanted the agent to be a back-end expert as a service."
This observation led to an important insight: If SoupBase or Stripe could generate more context configurations once, every developer using these services could benefit.
"Karo Powers officially formalizes that – things that people, just cutting-edge people, were doing – and allows anyone to have that kind of skill," Singh said.
Why dynamic loading pulses are fine-tuning for most AI coding use cases
The announcement also positions Caro Powers as a more economical alternative to fine-tuning, the process of training an AI model on specialized data to improve its performance in specific domains.
"It’s very cheap," said Singh, when asked how the options compare to fine-tuning. "Fine tuning is very expensive, and you can’t fix most Frontier models."
This is an important point. The most competent AI model Anthropicfor , for , for , . Open Eyeand Google Usually there are "closed source," Meaning developers cannot modify their base training. They can only influence the behavior of models through the signals and context they provide.
"Most people are already using powerful models like Sonnet 4.5 or Opus 4.5," Singh said. "What these models need is a point in the right direction."
Dynamic loading mechanism also reduces running costs. Because options are only activated when relevant, developers aren’t paying for token usage on tools they aren’t currently using.
Where Caro Powers fits into Amazon’s big bet on autonomous AI agents
Kero Powers arrives as part of a broader push by AWS that the company says "Agentic AI" – Artificial intelligence systems that can operate autonomously for extended periods of time.
Earlier in Re:Invention, AWS announced three "Frontier Agent" Designed to work for hours or days without human intervention: KIRO Autonomous Agent for Software Development, AWS Security Agent, and AWS Devops Agent. These represent a different approach from the Kero powers – dealing with larger, vaguer problems rather than providing specialized skills for specific tasks.
Both approaches are complementary. Frontier agents handle complex, multi-day projects that require independent decision-making across multiple code bases. KeroPowers, in contrast, provides developers with precise, efficient tools for everyday development tasks where speed and token efficiency matter most.
The company is betting that developers need both ends of this spectrum to be productive.
What Crow Powers Reveals About the Future of AI-Assisted Software Development
The launch reflects a maturing market for AI development tools. GitHub Coplot, launched by Microsoft in 2021, introduced millions of developers to AI-assisted coding. Since then, a proliferation of tools—incl Cursorfor , for , for , . Clineand Claude Code – What is the competition for the attention of developers.
But as these tools have become more capable, they’ve also become more complex. Model context protocolwhich was open-sourced by Anthropic last year, created a standard for connecting AI agents to external services. It solved one problem by creating another: the context overload that Caro Powers is now addressing.
AWS is positioning itself as the company that understands production software development at scale. Singh emphasized that Amazon’s 20 years of experience running AWS, combined with its own massive internal software engineering organization, gives it unique insight into how developers actually work.
"It’s not something you’ll just use for your prototype or your toy application," Singh said about AWS’s AI development tools. "If you want to build production applications, there’s a lot of knowledge that we bring to AWS that applies here."
The road ahead for Kero strengths and cross-platform compatibility
AWS indicated that Cero Powers currently only works within it Kiro Idebut the company is headed for cross-competition with other AI development tools, including command-line interfaces, Cursorfor , for , for , . Clineand Claude Code. The company’s documentation describes a future where developers can "Generate electricity once, use it anywhere" – Although this vision is aspirational for now.
For technology partners launching Powers today, the appeal is straightforward: Instead of maintaining separate integration documents for every AI tool on the market, they can create a single Powers that works everywhere. As more AI coding assistants crowd the market, this kind of efficiency becomes increasingly valuable.
Caro Powers Available now For developers using KIRO IDE version 0.7 or later at no additional charge beyond a standard KIRO subscription.
The basic premise is a familiar one in computing history: that the winners in AI-assisted development will not be the tools that try to do everything at once, but are smart enough to know what to forget.