"Working with LangSmith had a significant positive impact on the pace and quality of our development and shipping experience."
LangSmith Sandboxes
Give agents a secure computer
Ephemeral, isolated sandboxes designed for agent-generated code. Spin up a microVM in a single SDK call and let agents run untrusted code without touching your infrastructure.
LangSmith powers top engineering teams
.svg)
LangSmith Sandboxes
Secure, scalable, and stateful environments for every agentic AI workload
Run untrusted code safely
Spin up ephemeral microVMs for agent-generated code. Agents get full root inside the sandbox — they can install packages, run Docker, and do anything on a Linux box. Your core infrastructure and other workloads stay completely untouched.
MicroVM isolation
Agent code runs in a dedicated VM, fully isolated from your infrastructure and from other sandboxes.
Auth Proxy for credentials
Inject credentials into outbound API requests from the sandbox, so agents can call external services without secrets ever touching the runtime.
Framework-agnostic SDK
Use the Python or TypeScript SDK with Deep Agents, LangGraph, any other framework, or no framework at all.
.avif)
The Sandbox Runtime Teams Trust
Over 100,000 organizations use LangSmith to build and ship production-ready AI agents — including teams that run agent-generated code safely with Sandboxes
50M+
LLM Calls Traced
1B+
Events Ingested per Day
100K+
Monthly active orgs in LangSmith SaaS
LangSmith Agent Engineering Platform
Observe, evaluate, and deploy agents with LangSmith. LangSmith is framework-agnostic: trace your preferred framework or integrate LangSmith with any agent stack using our Python, TypeScript, Go, or Java SDKs.
Surface and diagnose undetected issues autonomously to improve agents faster. LangSmith Engine clusters production failures into prioritized issues, finds the root cause in your traces and code, and proposes the fix for your review.
Read the announcement
From isolated environment to production workload
1
Create
Start an isolated sandbox from the Python or TypeScript SDK.
2
Execute
Run agent-generated code with microVM isolation and protected credentials.
3
Scale
Launch parallel, stateful environments for production workloads.
An agent runtime your security team will approve
Isolate untrusted workloads while retaining enterprise controls.
MicroVM isolation
Each workload runs in a dedicated environment isolated from your infrastructure.
Protected credentials
Auth Proxy keeps secrets outside the runtime while allowing approved API access.
Enterprise controls
Apply role-based access, project isolation, and organizational governance.
Guides
Resources to get you started
Deep-dive guides from the LangChain team on shipping agents that run code safely in production.
New Guide
The Agentic Operating Model
How leading enterprise teams are building, deploying, and scaling AI agents in production. A framework for aligning the people, process, and technology needed to ship reliable agent systems.
New Guide
Agents Need Their Own Computer. Here's How to Give Them One Safely.
How to give agents a real execution environment with safe isolation, strong controls, full observability, and fast iteration — without exposing your infrastructure.
Meet us at an upcoming event
Join the LangChain community to learn how teams are building production agents.
Customers
FAQs for LangSmith Sandboxes
See LangSmith Sandboxes in Action
Get a personalized demo of LangSmith Sandboxes and see how your team can run agent-generated code safely at scale.