AI agents work on file systems using standard tools to navigate directories and read file paths.
The problem is that object storage systems, especially Amazon S3, contain a lot of enterprise data. Object stores serve data through API calls, not file paths. Bridging this gap requires a separate file system layer alongside S3, replicated data, and synchronization pipelines to align the two.
TThe rise of agent AI is exacerbating this problem, and it’s affecting Amazon’s ability to do its jobs. Engineering teams at AWS using tools like Kiro and Claude Code kept running into the same problem: Agents defaulted to local file tools, but the data was in S3. Downloading it locally worked until the agent’s context window was compressed and the session state was lost.
Amazon’s answer is S3 Files, which installs any S3 bucket directly into an agent’s local environment with a single command. Data remains in S3 without requiring migration. Under the hood, AWS integrates Elastic File System (EFS) technology with S3, providing full file system semantics, not a workaround. S3 Files are now available in most AWS Regions.
"By making data instantly available in S3 as if it were part of the local file system, we’ve seen a really big speedup with the ability for things like Kiro and Claude Code to be able to work with that data." Andy Warfield, vice president and principal engineer at AWS, told VentureBeat.
The difference between file and object storage and why it matters
S3 is built for object-level persistence, scalability, and API-based access. These features have made it the default storage layer for enterprise data. But they also created a fundamental incompatibility with the file-based tools that developers and agents depend on.
"S3 is not a file system and lacks file semantics on many fronts," Warfield said. "You can’t do movement, atomic movement of an object, and S3 doesn’t actually have directories."
Previous attempts to bridge this gap relied on FUSE (File Systems in USERspace), a software layer that allowed developers to mount a custom filesystem in userspace without changing main storage. Tools such as AWS’s own Mount Point, Google’s gcsfuse, and Microsoft’s blobfuse2 all used FUSE-based drivers to expose their respective object stores as filesystems.
The problem, Warfield noted, is that these object stores still don’t have file systems. Those drivers either faked file behavior by stuffing extra metadata into buckets, which broke the object API view, or refused file operations that the object store couldn’t support.
S3 Files have a completely different architecture. By integrating AWS EFS (Elastic File System) technology directly into S3, it provides a full native file system layer while keeping S3 as the system of record. Both the file system API and the S3 object API remain available simultaneously against the same data.
How S3 Files accelerates agent AI
Prior to S3 Files, an agent working with object data had to be explicitly instructed to download the files before using the tools. This created a session state problem. As agents crowded context windows, records of what was downloaded locally were often lost.
"I had to remind the agent that the data was available locally," Warfield said.
Warfield went through a before and after for a common agent task involving log analysis. He explained that the developer uses Kiro or Claude Code to work with the log data, only in the object they have to tell the agent where the log files are located and go and download them. If the logs can be immediately mounted to the local file system, the developer can simply specify that the logs are on a specific path and the agent can immediately traverse them.
For multi-agent pipelines, multiple agents can access the same mounted bucket at the same time. AWS says that thousands of compute resources can simultaneously connect to a single S3 file system, with total read capacity reaching many terabytes per second — numbers that VentureBeat could not independently verify.
Shared state between agents works through standard file system conventions: subdirectories, log files, and shared project folders that can be read and written by any agent in the pipeline. Warfield described AWS engineering teams using this pattern internally, where agents enter investigation notes and task summaries into shared project folders.
For teams building RAG pipelines over shared agent content, S3 vectors — launched at AWS re:Invent in December 2024 — overlays for extended generation with similarity search and search against the same data.
What analysts say: it’s not just a better FUSE
AWS hosts S3 Files against FUSE-based file access from Azure Blob NFS and Google Cloud Storage FUSE. The meaningful difference for AI workloads isn’t primarily performance.
"S3 Files eliminates data confusion between object and file storage, making S3 a shared, low-latency workspace without copying data." Gartner analyst Jeff Vogel told VentureBeat. "The file system becomes a view, not just another set of data."
With FUSE-based approaches, each agent maintains its own local view of the data. When multiple agents are running concurrently, these views can potentially be synchronized.
"It eliminates a whole class of failure modes, including unexplained training/output errors caused by outdated metadata," Vogel said. "FUSE-based solutions eliminate complexity and problems for the user."
At the agent level, the effects go even further. The architectural argument is less important than what it reveals in practice.
"This is the missing link for an agent AI that thinks in terms of files, paths, and native scripts." IDC analyst Dave McCarthy told VentureBeat. "It allows an AI agent to treat an exabyte-scale bucket as its own local hard drive, allowing for a level of autonomous transaction speed previously filled with API overhead associated with approaches such as FUSE."
Beyond agent workflow, McCarthy sees S3 Files as a broader entry point for how enterprises use their data.
"The launch of S3 Files isn’t just S3 with a new interface; it’s removing the last point of friction between massive data lakes and autonomous AI," he said. "By combining file and object access with S3, they open the door to more use cases with less development."
What this means for businesses
This architecture is now unnecessary for enterprise teams that maintain a separate file system alongside S3 to support file-based applications or agent workloads.
The practical transition for enterprise teams integrating AI infrastructure on S3 is concrete: S3 stops being a destination for agent output and becomes the environment where agent work happens.
"All these API changes you see from storage groups come from first-hand work and customer experience using agents to work with data." Warfield said. "We’re really focused on removing any friction and improving that interaction as much as we can."
