Skip to main content
Agents overwhelmingly prefer to access and manipulate data from a file system interface. Archil makes it simple to give your agents file systems that scale infinitely, only charge for what you use, and provide a way to directly run commands without a sandbox. 
# Spin up a file system in milliseconds
$ npx disk create agent-disk
agent-disk: dsk-abc123

# Run commands directly on your file system
$ npx disk dsk-abc123 exec "ls"

​
Fully elastic storage

Agents need workspaces for data, and you don’t know whether they’ll need 1 MB or 1 GB ahead of time. Archil is fully elastic, so you don’t need to pre-provision an amount of capacity, and you only pay for what you use. For more details, see Metering.

​
Share storage across agents

Agents need the ability to work on the same source data from multiple conversations, and you need to debug that data locally. Archil file systems can be mounted on multiple servers for simultaneous usage. For more information, see Shared Disks.

​
Run commands on your file system

Agents need the ability to run code that they write in order to accomplish tasks. Archil file systems can directly run bash, python, or node code without the need for a separate sandboxing service. For more information, see Serverless Execution.

​
Synchronize to S3

Agents need to work with data that you already have, where it currently lives. Archil file systems can optionally synchronize to S3, GCS, R2, or any S3-compatible storage system. For more information, see S3 Object Storage.