Box

Box · Launch Video Breakdown: Hook, Pacing & Motion Design

Helping devs build the future of intelligent, content-driven apps

0:00 · The Hook · The Sandboxed Agent Data Problem
0:00 / 0:00
AI AgentsProductionSeptember 8, 2026

Problem Agitation

5.15

Cuts / min

5:38

Runtime

55K

Views

49

Likes

9

Reposts

5

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    The Sandboxed Agent Data Problem

    “Say you're building an agent that runs inside a sandbox. It needs to gather documents, run code and tools against them, create new files, and also hand the final work back to your team. Now the agent part - that's exciting. But, moving all that content in and out, that might become its own project. You know - with all the downloads, uploads, sync, logic, retries. That's exactly the problem that Box Mount is built to solve.”

    On screen
    Say you're building an agent that runs inside a sandbox.
    Camera
    Static medium close-up of presenter in studio with warm/blue rim lighting.
    Motion
    Direct-to-camera address paired with crisp subtitle popups.
  2. Product Reveal

    Introducing Box Mount

    “In short, Box Mount brings a Box folder into the sandbox as a POSIX-compatible workspace to the agent. To the agent, it just looks and feels like a normal filesystem, meaning it can use familiar libraries, command line tools, and agent frameworks. So beyond just retrieving some context, Box Mount is for workflows where documents really become part of the agent's working environment. What you get is infrastructure built for developers creating sandboxed agent workflows. So, to show you what it looks like, I built a small contract review app powered by Box Mount.”

    On screen
    In short, Box Mount brings a Box folder into the sandbox as a POSIX-compatible workspace to the agent.
    Camera
    Picture-in-picture presenter over split-screen showing Box web UI and E2B dashboard.
    Motion
    Smooth layout zoom and split comparison.
  3. Feature Teaser

    Seeding the Workspace

    “First off, I have an empty Box folder. And locally, I have a sample MSA. And alongside that, I also have the legal team's approved playbook. First, we're going to seed that workspace into Box. After that, that's going to create an E2B sandbox and bring the Box content into the agent's working environment. The agent's going to review the agreement against the playbook, then synchronize the finished review back to Box for the team to review. Let's check it out. We'll start with the setup check. This validates my Box and E2B access, the sample files, the Box Mount binary, and my selected AI provider. In this case, I'm going to be using an agent from OpenAI. Next let's go ahead and seed the workspace. This sample app uses Box Mount inside a temporary workspace to synchronize the agreement and playbook into Box. And just double check - here they are. The files are in Box and our workspace is ready.”

    On screen
    empty Box folder. npm run doctor npm run seed
    Camera
    Screen recordings transitioning between macOS Finder, Terminal CLI, and Box Web UI.
    Motion
    Terminal output stream capture with smooth zoom crops.
  4. Feature Teaser

    Agent Review & Bidirectional Sync

    “And finally, let's start the review. The app creates a fresh E2B sandbox, installs Box Mount, and mounts the Box folder into that environment. The agent then reads the agreement and playbook, creates a review memo, and saves it into the mounted workspace. Just to emphasize, for this run, I've configured an OpenAI agent inside E2B. If I didn't provide a key and just kept it blank, this app's going to use Box AI by default. Back in Box, we're assigned a review task. So essentially what we just saw is an agent doing the first pass on MSA, then handing the result back to a human. So on that report - we have the risks that it found, suggested changes, and also the questions that still need human judgment. And here's the same workspace inside E2B. To the agent, it just looks like a normal filesystem. Box Mount provides a workspace and keeps it synchronized with Box. So for example, let's go ahead and add a quick note in Box and save it. Then if we go back to E2B - there it is. Changes synchronize in both directions, so the agent can pick up where the human left off.”

    On screen
    npm run demo Review complete and synchronized to Box.
    Camera
    Terminal CLI to Box task review interface to live markdown preview.
    Motion
    Bidirectional sync demonstration highlighting instantaneous UI reflection.
  5. Feature Teaser

    Architecture & Code Deep Dive

    “Now the best part, because this runs in a sandbox, I can add other scripts, libraries, and command line tools. I could also swap in another agent framework or give multiple agents separate environments. Again, Box Mount provides a content workspace, but what runs inside it is always up to the developer. All right, now that we saw this in action, let's check out how the sample app is built under the hood. Here, the app creates an E2B sandbox and passes in credentials and configurations that are needed for this run. Basically, each agent gets an isolated environment where it can run its own code and tools. Next up, this runs the Box Mount command inside E2B and mounts a selected Box folder into the sandbox. This way, Box content becomes a familiar filesystem workspace for the agent. After that, the agent reads the agreement and playbook using normal Node file tools. The takeaway here is that developers don't need to build a separate download and upload step for every file the agent touches. And finally, we can always inspect the live mount at any time. Then when we're done, we can always run a teardown, which performs a final sync, unmounts the folder, and destroys the sandbox. Keep in mind that everything that's synchronized to Box stays there.”

    On screen
    export async function createDemoSandbox npm run status npm run teardown
    Camera
    IDE code view scrolling, switching back to CLI status inspection.
    Motion
    Syntax-highlighted code showcase with terminal status monitoring.
  6. Call to Action

    Request Private Preview Access

    “Box Mount is currently in private preview. If you're building or starting to explore agent workflows that run in sandboxes and work with Box content, we'd love to hear what you're building. Check out the link to sign up and request preview access. Until then, I'm Andrew - see you next time.”

    On screen
    Box Mount is currently in private preview.
    Camera
    Direct frontal medium close-up studio shot.
    Motion
    Clean studio lighting with dynamic on-screen subtitles.

Related AI Agents Product Launches

Explore all AI Agents launches →
OpenAI
@OpenAI
AI Agents
@OpenAI
OpenAI@OpenAI
Gopuff
@gopuff
AI Agents
@gopuff
Gopuff@gopuff
Elon Musk
@elonmusk
AI Agents
@elonmusk
Elon Musk@elonmusk