# Contents

Volumes: Share Large Files Across Sandboxes Instantly

Here’s a real-world problem we kept hearing: a user creates an agent that needs a large dataset to work. Every time it spins up a new sandbox, it has to download that dataset from S3. Slow, repetitive, and expensive.

So we fixed it.

With Daytona Volumes, agents can now mount shared file systems across sandboxes. Download the data once, mount it anywhere, and start working instantly.

How It Works

Volumes are FUSE-based mounts that appear as normal folders inside the Sandbox file system. They let agents:

  • Access large files instantly

  • Skip repetitive downloads

  • Share data across multiple Sandboxes

  • Persist files beyond a single task or session

Everything stored in a Volume lives in an S3-compatible object store. So the agent can download it once, and then reuse it across environments.

Example: One Volume, Many Sandboxes

1from daytona_sdk import Daytona, CreateSandboxParams, VolumeMount
2
3daytona = Daytona()
4
5# Create a new volume or fetch an existing one
6volume = daytona.volume.get("my-volume", create=True)
7
8# Mount that volume into a new sandbox
9mount_path = "/mnt/data"
10
11sandbox = daytona.create(
12 CreateSandboxParams(
13 language="python",
14 volumes=[VolumeMount(volumeId=volume.id, mountPath=mount_path)],
15 )
16)

The agent can now read or write to /mnt/data just like any normal folder. The data persists even after the sandbox is destroyed.

Works Across Many Tasks

Volumes are perfect for agents working with:

  • Large models or datasets

  • Configuration bundles

  • Output files that need to be reused or stored

  • Shared assets across environments

You can mount multiple volumes to a single sandbox, or mount the same volume to many sandboxes. Whatever the task, the data is right there when the agent needs it.

Clean Up When You're Done

1# Remove the sandbox, but keep the volume
2daytona.remove(sandbox)
3
4# Later, delete the volume if it's no longer needed
5daytona.volume.delete(volume)

TL;DR

Agents now have shared memory. Upload once. Mount everywhere. Persist forever.

Volumes are live in the Daytona SDK.

Tags::
  • Daytona SDK
  • Agent Workflows
  • Shared Storage
  • Sandbox Persistence
  • Volumes
  • AI Infra
  • Dataset Sharing
  • FUSE
  • Reproducible Environments