Cursor, the AI-powered code editor, has released a public beta of its TypeScript SDK, enabling developers to programmatically invoke the same coding agents that power its desktop app, CLI, and web interface. The SDK provides access to a secure runtime with sandboxed cloud VMs, subagents, hooks, and token-based pricing, allowing organizations to integrate AI coding capabilities into their own workflows, CI/CD pipelines, or products.
Instead of requiring developers to interact with an agent in an IDE, the Cursor SDK allows agents to be called from any TypeScript application with a few lines of code. A minimal example from Cursor's announcement shows how to create an agent and send a task:
import { Agent } from "@cursor/sdk";
const agent = await Agent.create({
apiKey: process.env.CURSOR_API_KEY!,
model: { id: "composer-2" },
local: { cwd: process.cwd() },
});
const run = await agent.send("Summarize what this repository does");
for await (const event of run.stream()) {
console.log(event);
}
The Agent.create() method accepts an apiKey, a model specification, and either a local or cloud configuration for execution environment. This marks a shift from treating AI coding tools as interactive assistants to deployable infrastructure.
Cursor emphasizes that building reliable coding agents requires significant engineering effort for secure sandboxing, state management, environment setup, and context handling. The SDK aims to remove this complexity so teams can focus on building useful agents rather than maintaining underlying infrastructure.