How StackAI Runs Enterprise AI Agents for Regulated Industries
.png)
StackAI is a no-code enterprise AI platform used by enterprise teams to build and deploy agentic workflows on sensitive data. Founded in 2023 by two MIT PhDs, Tony Rosinol and Bernard Aceituno, it lets users automate document- and process-heavy work by dragging and dropping nodes, plugging in any LLM, and connecting natively to enterprise systems like SharePoint and SAP.
Those workflows run code. Users trigger it, and so do the LLMs. In regulated industries, that code cannot run anywhere it can touch another customer's data or the host. As StackAI's customer workflows grew more sophisticated, the team hit a wall. The team needed secure, isolated execution baked into every workflow, at scale, and off-the-shelf options like REPL did not meet the bar.
The problem: building sandbox infrastructure is a full-time job
StackAI initially built its own sandbox system to support user-triggered and LLM-triggered code execution. It held up at small scale. As demand grew, the team found itself maintaining provisioning, network isolation, session persistence, and enterprise-grade access control in-house. None of it scaled cleanly, and all of it pulled engineers off the product - especially as adoption rapidly grew.
"We built our own code sandbox VM, and it neither scaled well nor had all the features we needed. We couldn’t afford the time to take away from building revenue-generating features for 5-6 of our platform engineers to build and maintain the infrastructure".
The team’s requirements for an ideal solution were specific:
- Real isolation by default. StackAI's customers operate in banking, defense, and healthcare. Agent-generated code has to run in an environment hardened against escape, not a shared container. E2B runs every sandbox as an isolated VM.
- Fast, easy provisioning. A sandbox spins up per execution. If that step is slow or fiddly, it shows up in every workflow run.
- Scale. Hundreds of agent workflows in production per customer, with headroom to grow.
"What stood out initially was how fast E2B was to get going from running a few sandboxes for evaluation to hundreds of thousands of sandboxes. It just worked."
The integration: three capabilities, one runtime
After the initial evaluation, the StackAI team quickly scaled E2B usage across three core use cases.
Custom Code. Users write and execute Python or JavaScript directly inside agent workflows through the Code Node. E2B handles the sandboxing, so every execution runs in a secure, isolated environment regardless of workflow complexity.
Computer Use and RPA. Users first map out what a successful session looks like by recording and identifying the right steps. StackAI then replays those steps within E2B sandboxes, enabling agents to interact with desktop environments and enterprise software securely and reliably.
Browser Use. Agents that navigate the web, scrape data, or drive legacy and browser-based tools run those sessions inside E2B, keeping every session isolated and secure.
The throughline across all three is the same: isolation that customers in regulated industries can trust, provisioning the team does not have to manage, and performance that holds at scale.
"With E2B, it's really easy to provision sandboxes. They're secure by default, highly capable, and very performant. E2B is one of the only providers that delivers the enterprise security we need at the scale and throughput that helps us create experiences that delight our end users and solve their problems."
What's next: a sandbox for every agent
StackAI's roadmap treats the sandbox as core infrastructure, not a feature. The plan is for every agent to get its own sandbox. Not just for code execution, but as a persistent environment with its own datastore: a place to keep data, run processes, and maintain state across sessions.
That bet depends on the runtime keeping up.
"Sandboxes are a key primitive for agents going forward. For us to keep growing as a company, we need a partner that can grow with us - we’re happy to have found one in E2B."
Follow StackAI for more:
Secure AI Sandbox
