Had a blast talking “company brain” at @lightsagehq's demo night yesterday!
Huge thanks to @jun_liang_sf and @sean__er for having me. And congrats on the funding! Great guys building some seriously cool stuff 🚀
We're going to dive into how enterprise app dev is moving from SQL + web to unstructured data + AI ⚡️
If you're in SF and want to join us, DM me and I'll hook you up with a pass 💙
Box AI Extract pulls structured fields from documents, and @OpenAI Decisions API routes them to the right workflow.
In this demo, a COI arrives in Box through a File Request. Box AI extracts coverage fields from both the upload and the project requirements. The Decisions API receives that evidence and returns a typed route: request endorsement, specialist review, or ready for human review.
A Box webhook triggers the whole process on document upload. The app creates a review task with context, adds the right collaborator, and writes extracted values and decision confidence back to the file as metadata.
Box is where the documents arrive and where people review them.
Watch👇
We are 3 weeks out from the first ever BoxWorks Dev Summit!
We'll be exploring how AI is unlocked by unstructured data with:
- @ivanburazin / @daytonaio
- @charlierguo / @OpenAIDevs
- Joey Conway / @nvidia
- @sobtiankit / @getpostman
Plus, hands-on workshops from @awscloud , NVIDIA, OpenAI and @Docker.
If you're in the SF Bay Area, we have a limited number of free passes for devs, use the link below to RSVP!
https://t.co/9pNmMCTMXn
Another spin on running agents in the open source MXC sandbox from @Microsoft, this time with a Box Mount-backed workspace. Fun to see all the pieces working together 🤙
🚀@Microsoft Execution Containers (MXC) provide policy-driven, sandboxed execution across Windows, Linux, and macOS. Box Mount runs inside that sandbox and maps a Box folder to a standard POSIX path, with two-way sync and Box permissions and governance applying automatically.
In this demo, the agent reconciles fictional invoices, purchase orders, and payments using deterministic tools. It identifies six exceptions including duplicate records, a missing PO, an over-limit invoice, and unpaid invoices. The finished Markdown review and Excel exception workbook sync back to Box through Box Mount.
MXC is the isolated execution layer, and Box Mount is the governed content layer. The result lands in Box, ready for human review.
Demo ↓
Box Mount lets developers mount Box as a filesystem directly in any agent sandbox, so agents can work with files and data just like a person would.
In this article, we show how it works with @DigitalOcean, @E2B, @OpenAI, and @vercel. https://t.co/JXeN5xhNb7
A sandbox agent shouldn’t need custom download and upload steps for every document it touches.
With Box Mount, Box content appears at an ordinary path inside a @digitalocean sandbox. The agent works with the files there, and its output syncs back to Box for review.
Demo 👇
We built Box Mount to allow developers to mount Box as a file system directly in any agent sandbox. As agents do more complex work, they need to be able to work with files and data just like a person would. The future is headless.
🚀@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.
Giving an agent a job doesn’t mean handing it the keys to everything 👀
MXC + Box Drive in action: approved files into the sandbox, agent does its thing, results back in Box for 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 ↓
Loved stopping by SF @Techweek_'s Zero Human Startup Hackathon today!
Spoke about giving agents a company brain with @Box. Somewhere to find the docs, decisions, and context behind the work you’re asking them to do 🤠
Great turnout at the @Box booth at @WeAreDevs this afternoon!
Everyone was definitely here to learn about AI agents and filesystems. The raffle was purely coincidental 😬
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: https://t.co/SjG8fIacqt
All that time reading about AI finally has a purpose: winning trivia 🤠
We’re co-hosting a @WeAreDevs pre-party at Guildhouse in San Jose. AI trivia, board games, tacos, and a little friendly competition. Come hang!
https://t.co/80cUHPZ3QD
Testing Jev since yesterday. Love seeing Khaneman's system 1 in practice. Such a great developer experience, elegant solution, and amazing launch. Well done @typesafeai! ✨ Stay tuned for a full article