Amazon Web Services on Tuesday announced a brand new class of artificial intelligence systems called "frontier agents" that may work autonomously for hours and even days without human intervention, representing one of the vital ambitious attempts yet to automate the total software development lifecycle.
The announcement, made during AWS CEO Matt Garman's keynote address at the corporate's annual re:Invent conference, introduces three specialized AI agents designed to act as virtual team members: Kiro autonomous agent for software development, AWS Security Agent for application security, and AWS DevOps Agent for IT operations.
The move signals Amazon's intent to leap ahead within the intensifying competition to construct AI systems able to performing complex, multi-step tasks that currently require teams of expert engineers.
"We see frontier agents as a totally latest class of agents," said Deepak Singh, vice chairman of developer agents and experiences at Amazon, in an interview ahead of the announcement. "They're fundamentally designed to work for hours and days. You're not giving them an issue that you simply want finished in the following five minutes. You're giving them complex challenges that they could need to take into consideration, try different solutions, and get to the correct conclusion — they usually should try this without intervention."
Why Amazon believes its latest agents leave existing AI coding tools behind
The frontier agents differ from existing AI coding assistants like GitHub Copilot or Amazon's own CodeWhisperer in several fundamental ways.
Current AI coding tools, while powerful, require engineers to drive every interaction. Developers must write prompts, provide context, and manually coordinate work across different code repositories. When switching between tasks, the AI loses context and must start fresh.
The brand new frontier agents, in contrast, maintain persistent memory across sessions and repeatedly learn from a company's codebase, documentation, and team communications. They’ll independently determine which code repositories require changes, work on multiple files concurrently, and coordinate complex transformations spanning dozens of microservices.
"With a current agent, you’d go microservice by microservice, making changes separately, and every change can be a distinct session with no shared context," Singh explained. "With a frontier agent, you say, 'I want to unravel this broad problem.' You point it to the correct application, and it decides which repos need changes."
The agents exhibit three defining characteristics that AWS believes set them apart: autonomy in decision-making, the power to scale by spawning multiple agents to work on different features of an issue concurrently, and the capability to operate independently for prolonged periods.
"A frontier agent can resolve to spin up 10 versions of itself, all working on different parts of the issue directly," Singh said.
How each of the three frontier agents tackles a distinct phase of development
Kiro autonomous agent serves as a virtual developer that maintains context across coding sessions and learns from a company's pull requests, code reviews, and technical discussions. Teams can connect it to GitHub, Jira, Slack, and internal documentation systems. The agent then acts like a teammate, accepting task assignments and dealing independently until it either completes the work or requires human guidance.
AWS Security Agent embeds security expertise throughout the event process, routinely reviewing design documents and scanning pull requests against organizational security requirements. Perhaps most importantly, it transforms penetration testing from a weeks-long manual process into an on-demand capability that completes in hours.
SmugMug, a photograph hosting platform, has already deployed the safety agent. "AWS Security Agent helped catch a business logic bug that no existing tools would have caught, exposing information improperly," said Andres Ruiz, staff software engineer at the corporate. "To some other tool, this may have been invisible. But the power for Security Agent to contextualize the data, parse the API response, and find the unexpected information there represents a breakthrough in automated security testing."
AWS DevOps Agent functions as an always-on operations team member, responding immediately to incidents and using its accrued knowledge to discover root causes. It connects to observability tools including Amazon CloudWatch, Datadog, Dynatrace, Recent Relic, and Splunk, together with runbooks and deployment pipelines.
Commonwealth Bank of Australia tested the DevOps agent by replicating a posh network and identity management issue that typically requires hours for knowledgeable engineers to diagnose. The agent identified the basis cause in under quarter-hour.
"AWS DevOps Agent thinks and acts like a seasoned DevOps engineer, helping our engineers construct a banking infrastructure that's faster, more resilient, and designed to deliver higher experiences for our customers," said Jason Sandry, head of cloud services at Commonwealth Bank.
Amazon makes its case against Google and Microsoft within the AI coding wars
The announcement arrives amid a fierce battle amongst technology giants to dominate the emerging marketplace for AI-powered development tools. Google has made significant noise in recent weeks with its own AI coding capabilities, while Microsoft continues to advance GitHub Copilot and its broader AI development toolkit.
Singh argued that AWS holds distinct benefits rooted in the corporate's 20-year history operating cloud infrastructure and Amazon's own massive software engineering organization.
"AWS has been the cloud of selection for 20 years, so now we have twenty years of data constructing and running it, and dealing with customers who've been constructing and running applications on it," Singh said. "The learnings from operating AWS, the knowledge our customers have, the experience we've built using these tools ourselves every single day to construct real-world applications—all of that’s embodied in these frontier agents."
He drew a distinction between tools suitable for prototypes versus production systems. "There's quite a lot of things on the market that you could use to construct your prototype or your toy application. But when you must construct production applications, there's quite a lot of knowledge that we usher in as AWS that apply here."
The safeguards Amazon built to maintain autonomous agents from going rogue
The prospect of AI systems operating autonomously for days raises immediate questions on what happens after they go off course. Singh described multiple safeguards built into the system.
All learnings accrued by the agents are logged and visual, allowing engineers to know what knowledge influences the agent's decisions. Teams may even remove specific learnings in the event that they discover the agent has absorbed misinformation from team communications.
"You’ll be able to go in and even redact that from its knowledge like, 'No, we don't want you to ever use this data,'" Singh said. "You’ll be able to take a look at the knowledge prefer it's almost—it's like your neurons inside your brain. You’ll be able to disconnect some."
Engineers may also monitor agent activity in real-time and intervene when obligatory, either redirecting the agent or taking on entirely. Most critically, the agents never commit code on to production systems. That responsibility stays with human engineers.
"These agents are never going to ascertain the code into production. That continues to be the human's responsibility," Singh emphasized. "You’re still, as an engineer, answerable for the code you're checking in, whether it's generated by you or by an agent working autonomously."
What frontier agents mean for the longer term of software engineering jobs
The announcement inevitably raises concerns concerning the impact on software engineering jobs. Singh pushed back against the notion that frontier agents will replace developers, framing them as a substitute as tools that amplify human capabilities.
"Software engineering is craft. What's changing isn’t, 'Hey, agents are doing all of the work.' The craft of software engineering is changing—how you employ agents, how do you arrange your code base, how do you arrange your prompts, how do you arrange your rules, how do you arrange your knowledge bases in order that agents could be effective," he said.
Singh noted that senior engineers who had drifted away from hands-on coding at the moment are writing more code than ever. "It's actually easier for them to turn out to be software engineers," he said.
He pointed to an internal example where a team accomplished a project in 78 days that might have taken 18 months using traditional practices. "Because they were in a position to use AI. And the thing that made it work was not only the undeniable fact that they were using AI, but how they organized and arrange their practices of how they built that software were maximized around that."
How Amazon plans to make AI-generated code more trustworthy over time
Singh outlined several areas where frontier agents will evolve over the approaching years. Multi-agent architectures, where systems of specialised agents coordinate to unravel complex problems, represent a serious frontier. So does the combination of formal verification techniques to extend confidence in AI-generated code.
AWS recently introduced property-based testing in Kiro, which uses automated reasoning to extract testable properties from specifications and generate hundreds of test scenarios routinely.
"If you may have a shopping cart application, every way an order could be canceled, and the way it may be canceled, and the best way refunds are handled in Germany versus the US—in case you're writing a unit test, perhaps two, Germany and US, but now, because you may have this property-based testing approach, your agent can create a scenario for each country you use in and test all of them routinely for you," Singh explained.
Constructing trust in autonomous systems stays the central challenge. "Without delay you continue to require tons of human guardrails at every step to make certain that the correct thing happens. And as we recuperate at these techniques, you’ll use less and fewer, and also you'll have the opportunity to trust the agents so much more," he said.
Amazon's greater bet on autonomous AI stretches far beyond writing code
The frontier agents announcement arrived alongside a cascade of other news at re:Invent 2025. AWS kicked off the conference with major announcements on agentic AI capabilities, customer support innovations, and multicloud networking.
Amazon expanded its Nova portfolio with 4 latest models delivering industry-leading price-performance across reasoning, multimodal processing, conversational AI, code generation, and agentic tasks. Nova Forge pioneers "open training," giving organizations access to pre-trained model checkpoints and the power to mix proprietary data with Amazon Nova-curated datasets.
AWS also added 18 latest open weight models to Amazon Bedrock, reinforcing its commitment to offering a broad number of fully managed models from leading AI providers. The launch includes latest models from Mistral AI, Google's Gemma 3, MiniMax's M2, NVIDIA's Nemotron, and OpenAI's GPT OSS Safeguard.
On the infrastructure side, Amazon EC2 Trn3 UltraServers, powered by AWS's first 3nm AI chip, pack as much as 144 Trainium3 chips right into a single integrated system, delivering as much as 4.4x more compute performance and 4x greater energy efficiency than the previous generation. AWS AI Factories provides enterprises and government organizations with dedicated AWS AI infrastructure deployed in their very own data centers, combining NVIDIA GPUs, Trainium chips, AWS networking, and AI services like Amazon Bedrock and SageMaker AI.
All three frontier agents launched in preview on Tuesday. Pricing will likely be announced when the services reach general availability.
Singh made clear the corporate sees applications far beyond coding. "These are the primary frontier agents we’re releasing, they usually're within the software development lifecycle," he said. "The issues and use cases for frontier agents—these agents which can be long running, able to autonomy, pondering, all the time learning and improving—could be applied to many, many domains."
Amazon, in any case, operates satellite networks, runs robotics warehouses, and manages certainly one of the world's largest e-commerce platforms. If autonomous agents can learn to write down code on their very own, the corporate is betting they’ll eventually learn to do absolutely anything else.
