Case Study

How LangChain Found a Trusted Partner for Their Sandbox Needs

Case Study

How LangChain Found a Trusted Partner for Their Sandbox Needs

4000

Sandboxes created per month with Daytona

660+

Hours of sandbox runtime per month

20+

Minute sessions fully supported with Daytona

LangChain is the leading open-source framework for building LLM-powered apps, helping companies like Klarna, Morningstar, and Clay launch reliable, scalable agent workflows faster.

Headquarters

San Francisco, CA

Industry

Technology

Department

Artificial Intelligence Engineering

Key Features

Sandbox Creation Speed Sandbox Statefulness Long‑Running Sandboxes Infrastructure Scale

Learn how the leading open-source agent engineering platform partnered with Daytona to ensure secure, instant, and persistent sandboxes for their production-ready coding agent.

Our mission is to be the one place developers go for all their agent needs. Daytona is the only tech partner that could support our sandbox needs and make this possible.

Brace Sproul

Head of Applied AI at LangChain

01 -- CHALLENGE

Deploying Open-Source Coding Agents Called for a Reliable Sandbox Infrastructure Partner


As a leader in open‑source agent development, LangChain’s bread and butter is building, orchestrating, and deploying agents seamlessly for developers and non‑technical users alike. To advance this mission, Brace Sproul, Head of Applied AI, and his team built a cloud‑hosted coding agent that runs autonomously end to end. To open-source it reliably, they needed to nail one thing: their sandbox partner.

For the agent to run safely across thousands of users and complex use cases, Brace needed secure, production-level sandboxes fast and at scale. But he and his team knew from experience that building internal infrastructure drained valuable resources and engineering time. That effort was better spent enriching libraries, enhancing evals, and debugging products. So they started looking for a partner who offered a solution out of the box.

The requirements weren’t trivial. Each sandbox needed bulletproof isolation to limit access to sensitive data and fine-grained permissions to ensure protection from external attacks. And instant spin‑up wasn't just a nice‑to‑have. Any delays could impact user engagement and break agent workflows.

Because coding workflows require persistence, their partner’s sandboxes needed to be stateful. “Our agent installs dependencies, compiles code, and edits code between steps,” explains Brace. “The sandbox has to run for this entire session, keeping the repo and all intermediate artifacts alive.”

To avoid low-on-space failures for each coding step and repo, each sandbox also had to be the right size. That called for a partner who could provide configurable disk capacity and adequate compute, so compile and test phases ran efficiently and didn’t crash.

What mattered to Brace wasn't just infrastructure, however. Running agents at scale meant dealing with unexpected production issues like networking timeouts and disk constraints. Even if the solution met every technical spec on paper, responsive, hands-on support was non-negotiable.

“Support directly impacts our development velocity,” shares Brace. “An issue that takes a week to resolve with one provider, can be fixed in hours with a partner who jumps in to help.” That difference ensures users get reliable features on time.

Soon, LangChain found a popular AI sandbox provider that promised to deliver results. At first, the platform “ran well enough.” Yet, as LangChain’s team hit hard limits on disk size, compute resources, and network access, workflow bottlenecks and errors stalled development. 

“I expected it to work out of the box, and instead, they blocked us for days,” says Brace. “When we reached out for help, they were very slow to respond.”

That’s when Brace found Daytona. Within hours of their first call, Daytona’s team had accessed LangChain’s open-source repo and migrated the entire codebase to their platform, delivering a working solution in just a few hours. That responsiveness and technical capability convinced Brace that he’d found the perfect partner for LangChain’s needs.

While building our coding agent, we hit limitations around our sandboxing. Daytona jumped in, contributed a working PR within hours, and fully unblocked us. They pulled a Collison.

Harrison Chase

CEO & Founder of LangChain

02 -- SOLUTION

A Responsive, Hands-On Partner That Delivers Secure, Persistent, and Scalable Sandboxes


Daytona was the infrastructure partner LangChain needed to open-source their agent and scale it to thousands of users. In just a few hours, Daytona started provisioning secure, reliable sandboxes, handling all edge cases and security complexities without any lift from Brace or his team.

This frictionless experience extends to the platform itself. For instance, rather than simply isolating every agent, it delivers granular control out of the box. Brace and his team fine‑tune access to specify which users and services may call into the sandbox, and the capabilities running agents have once inside.

With these granular permissions, LangChain prevents unauthorized activity while maintaining flexibility for complex, legitimate use cases. This control keeps security risks low and developer productivity high, delivering users the right resources without permission errors or lengthy access approvals.

Daytona provisions each environment in less than 100ms, ensuring a seamless user experience with minimal latency. And since sandboxes are built for persistent, long-running tasks, LangChain’s agent can install dependencies, modify files, and iterate on solutions without interruption. LangChain’s community contributors and users now enjoy reliable, long-duration agent workflows, making it easier to build, debug, and launch projects. The result is faster development that reduces operational costs and increases throughput.

Thanks to this reliable setup, LangChain allocates engineering hours and internal resources to the tasks that matter most: enhancing LangChain’s tools and libraries to facilitate agent testing, evals, and deployment. Because Brace and his team also configure details like CPU, RAM, and disk size limits, they optimize resource utilization without impacting performance or scalability.

These technical specs check every box on LangChain’s list. But what truly impressed Brace was Daytona's responsive assistance. As their cloud runtime platform hums in the background, their team stays hands-on through regular check-ins and swift online support.

For example, during one of their conversations, Brace and his team shared that they needed a reliable way to handle private repo access. Daytona instantly set them up with “the perfect solution,” ensuring authentication occurs outside the sandbox. Now, agents push and pull code without the risk of exfiltrating access tokens, supporting seamless data protection, more ambitious agentic workflows, and faster releases that drive innovation and agility.

“Whenever we ask the Daytona team a question, they reply quickly and walk us through our situation, even though we are an ocean apart,” Brace highlights. “The fact that they’re so engaged with our product development accelerates our development and feature launches.”

And the team doesn’t stop at attentive support. They continuously evolve the platform based on LangChain’s feedback. By actively seeking input and implementing improvements quickly, Daytona keeps LangChain focused on their core projects and pushes them further ahead in the fast-moving AI space.

The fact that Daytona is always so engaged and responsive is one of the main reasons why we chose them. Working with them has been seamless.

Brace Sproul

Head of Applied AI at LangChain

03 -- RESULT

LangChain Spins Up 4000 Sandboxes Every Month with Daytona

With Daytona as their infrastructure partner, LangChain successfully open-sourced their coding agent. Their team now has the foundation to rapidly deliver new features to their community and empowers a growing base of developers to build with reliable coding AI. 

At every step, they know that Daytona has their back. Their team is always ready to jump in with solutions, answer questions quickly, and keep development moving.

  • 4000 sandboxes created per month with Daytona

  • 660+ hours of sandbox runtime per month

  • 10-to-30-minute sessions fully supported with Daytona

  • <100ms sandbox creation time

Moving forward, Brace and his team will continue to use Daytona’s runtime platform to ensure agent enhancements run smoothly and securely for thousands of users. 

If anybody asked me which runtime platform to use, I'm going to 100% say Daytona every time. With their SDK, we never have to worry about an agent leaking or taking actions that it shouldn't. I haven't seen this anywhere else.

Brace Sproul

Head of Applied AI at LangChain

NEWSLETTER
Subscribe to DotFiles Insider, a fortnightly newsletter for developers covering stories, techniques, guides and the latest product innovations.
NEWSLETTER
Subscribe to DotFiles Insider, a fortnightly newsletter for developers covering stories, techniques, guides and the latest product innovations.
NEWSLETTER
Subscribe to DotFiles Insider, a fortnightly newsletter for developers covering stories, techniques, guides and the latest product innovations.