# Introducing Run SDK: secure eval for your agents

DevFeed: [Introducing Run SDK: secure eval for your agents](<https://devfeed.tech/articles/introducing-run-sdk-secure-eval-for-your-agents-763.md>)

Original publisher: [Read original article](<https://vercel.com/blog/introducing-run>)

Author: Aayush Kapoor

Published: 2026-08-25T04:00:00Z

Content type: release

Language: en

Sources: [Vercel News](<https://devfeed.tech/sources/vercel-news.md>)

Topics: [AI Bots](<https://devfeed.tech/topics/ai-bots.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [authentication](<https://devfeed.tech/tags/authentication.md>), [javascript](<https://devfeed.tech/tags/javascript.md>), [node-js](<https://devfeed.tech/tags/node-js.md>), [sandbox](<https://devfeed.tech/tags/sandbox.md>), [secrets](<https://devfeed.tech/tags/secrets.md>), [tools](<https://devfeed.tech/tags/tools.md>), [typescript](<https://devfeed.tech/tags/typescript.md>)

## AI overview

Vercel releases the Run SDK for executing untrusted JavaScript and TypeScript in an isolated QuickJS worker context. Applications expose selected host functions while keeping credentials and service clients outside the sandbox, and can pause execution for authentication or human approval before resuming work.

## Source excerpt

Agents increasingly write TypeScript programs to coordinate tools and process their results. Once those programs touch real applications, some steps require authentication, while others need human approval. Executing that code with eval gives it the same access as the application around it, including its secrets and internal services, and leaves no durable way to pause at those boundaries. Today, we're releasing the Run SDK, a package for executing untrusted JavaScript and TypeScript without giving it direct access to your application or system. Applications expose narrow host functions and can interrupt execution for authentication or human-in-the-loop approval. The program resumes after a decision without repeating completed work. A small interface to the host The Run SDK evaluates JavaScript or type-stripped TypeScript in a fresh QuickJS context inside a worker thread, with no direct route to Node.js or the network. The application exposes selected operations through hostFunctions. These are regular functions that become callable globals inside the sandbox: Here, the generated program knows about store.listOrders(). The database client and its credentials remain in the application. Calls cross the sandbox boundary through serialization. A host function may return a promise, so existing service clients can sit behind this interface without being passed into the sandbox. You can try this in the playground. Code you run there can only reach the host functions on the page. Code mode in practice The Run SDK is the internal module powering code mode tool execution in the AI SDK. Giving an agent a program changes the unit of work. One model response can describe the calls and the logic connecting them: The two requests happen concurrently, and invoice filtering stays local to the program. Only the useful result returns to the application. This is a good fit for agents that work across several internal services. A research agent can combine search results before answerin