Andrew Wong

🚀@vercel Sandbox gives AI agents an on-demand Linux microVM with its own filesystem, network, and process space. Box Mount runs inside that sandbox and maps a Box folder to a standard POSIX path, with 2-way sync and Box permissions and governance applying automatically.

In this demo, the agent analyzes 12 security questionnaire items against 3 approved policy documents. 10 answers pass with sourced evidence. 2 are flagged for human review because the approved docs don't support those claims. The finished Excel workbook and review list sync back to Box through Box Mount.

Vercel Sandbox is the isolated execution layer, and Box Mount is the governed content layer. The result lands in Box, ready for human review.

🚀@vercel Sandbox gives AI agents an on-demand Linux microVM with its own filesystem, network, and process space. Box Mount runs inside that sandbox and maps a Box folder to a standard POSIX path, with 2-way sync and Box permissions and governance applying automatically.

In this demo, the agent analyzes 12 security questionnaire items against 3 approved policy documents. 10 answers pass with sourced evidence. 2 are flagged for human review because the approved docs don't support those claims. The finished Excel workbook and review list sync back to Box through Box Mount.

Vercel Sandbox is the isolated execution layer, and Box Mount is the governed content layer. The result lands in Box, ready for human review.

Microsoft MXC gives AI agents a controlled environment to run code, with policies that define which files they can read, where they can write, and what they can access. This demo shows MXC running natively on @Windows, connected to governed business content in Box through Box Drive.

Only approved files are staged into MXC before the agent runs. Box Drive makes the content accessible through the desktop filesystem. When the work is finished, results return to Box for team review.

MXC enforces what the agent can do. Box provides secure, governed content management and brings the output back into the team's workflow. Together, they connect useful agent work with enterprise controls.

Demo below ↓

Microsoft MXC gives AI agents a controlled environment to run code, with policies that define which files they can read, where they can write, and what they can access. This demo shows MXC running natively on @Windows, connected to governed business content in Box through Box Drive.

Only approved files are staged into MXC before the agent runs. Box Drive makes the content accessible through the desktop filesystem. When the work is finished, results return to Box for team review.

MXC enforces what the agent can do. Box provides secure, governed content management and brings the output back into the team's workflow. Together, they connect useful agent work with enterprise controls.

Demo below ↓

New this year, join us for the first ever Dev Summit at BoxWorks. If you’re building agentic and content-driven enterprise applications and workflows you won’t want to miss this!

2 days of technical sessions and hands-on workshops covering everything from the latest in agentic sandboxes to file systems for agents. Speakers from @nvidia, @OpenAI, and Box and more coming soon! November 5-6 in San Francisco.

Register here: boxworks.box.com/series/boxwork…

Box and @OpenAI are partnering to bring governed enterprise content to AI-powered legal work with Astra for Law and @ChatGPT. Legal skills built on Box and ChatGPT help law firms turn their content into practical, secure workflows without weakening the controls around it.

In this demo, a legal professional uses the new Outside Counsel Guidelines skill to review an incoming set of OCGs directly from Box. ChatGPT flags associated risks and leaves comments for human review, all without the content leaving Box.

🚀@vercel's’ fx is a tiny open-source coding agent built to take action directly in a developer’s workspace. This demo extends that idea by giving fx access to the security policies, requirements, and release docs stored in Box.

Using the Box CLI, fx retrieves an approved policy, compares it with local code, edits the files, adds missing test coverage, and prepares release evidence without being told which lines to change. It asks for human approval before uploading anything, then verifies every file after it lands in Box.

Git stays where the code lives. Box stays where the approved docs and release proof live. fx connects the work across both.

Demo below ↓

Box + @Snowflake turn financial documents into queryable history, without losing the path back to the source.

Most financial research workflows follow the same pattern: extract details from documents, write a summary, copy numbers into another system, and hope everything can be retraced when someone asks where a figure came from.

This Python demo closes that loop. Box Extract transforms PDFs into structured financial data while preserving confidence scores and page references. Box Doc Gen produces a readable, sourced memo. Snowflake stores common fields alongside the complete JSON payload, creating queryable records that can be compared across quarters or companies.

The result is unstructured documents become useful data while every number remains connected to its source.

🆕 @OpenAIDevs Agents API gives developers a hosted sandbox where agents run tools, coordinate work, and handle complex tasks. Box Mount brings enterprise content directly into that sandbox as files the agent can read, reason across, and produce new work from, using normal shell commands and file paths while Box Mount handles two-way sync automatically.

In our demo, a lead agent mounts a deal room from Box, reads five source files, launches three specialist agents in parallel, and writes four reports back through Box Mount with an initial verdict. When a customer revises the MSA in Box, Box Mount syncs the new version into the same sandbox and the agents reassess automatically, refreshing all four reports while preserving the revised MSA as version two in Box.

The result is a shared workspace where people and agents work on the same governed Box content without building custom file-transfer logic, with permissions, versions, and audit history preserved throughout.

Box Mount is in private preview. Watch👇

Proud to have my contract clause extraction agent template with @Box and @asana published to @claudeai Managed Agents ✨

We’re introducing Box Mount.

Mount a Box folder into a sandbox like @e2b and keep it in two-way sync. The agent reads and writes Box content with standard file operations as if they were local. Every operation flows through the Box API so governance applies automatically, no custom policy code required.

2 commands to get started:
agent-mount config
agent-mount mount "/mnt/box" "<box-id>"

Developer Preview is now open.

OpenClaw 2.0 is out and /loop is now part of the stable release.

Connected to Box, here's what that looks like. One prompt connects @openclaw to an M&A deal room via the Box CLI and starts a recurring check. When a signed customer renewal lands in the folder, OpenClaw catches it on the next pass, connects it to the existing risk register, updates the relevant deal context, and saves the update as a new version of the same file in Box, with the agent identity preserved throughout.

Start it with one prompt. Stop it with /loop stop. OpenClaw keeps up with the work as it changes, and Box stays the source of truth.

Anthropic’s @Claude Fable 5.1 is built for multi-step analytical enterprise work. We ran it through Box's Complex Work Eval against Fable 5 :

→ Overall: 72% vs Fable 5's 65% (+7pp)
→ Financial Services: +19pp over Fable 5
→ Technology: +37pp over Fable 5
→ Public Sector: +16pp over Fable 5
→ Data Analysis: +12pp over Fable 5

A clear step up on the quantitative, multi-step work that drives real decisions. Fable 5.1 is 23% faster and uses 25% fewer total tokens than Fable 5. More accurate and more efficient at the same time.

Most teams governing AI agents focus on what they can access. The harder problem is what they can do with it.

Box MCP Server admin controls let admins govern tool access enterprise-wide or per AI integration, set read-only or read-and-write permissions, and configure down to the individual tool level. When a tool is disabled, agents can no longer discover or invoke it. Enforced at runtime, not just at setup.