Product Updates

Cursor brings Cloud Agent execution to self-hosted machines

Cursor’s September 2026 self-hosted machines release keeps Cloud Agent tool execution, code, build outputs, and secrets inside a team’s own network, with dynamic worker pools and sandbox options.

By Authority AI Tools Editorial Team2026-09-0210 min read
Last reviewed: 2026-09-02
AATET
Authority AI Tools Editorial Team

Editorial Team

The Authority AI Tools editorial team maintains this directory using vendor documentation, dated source checks, product changelogs, and clearly identified hands-on observations where available.

Cursor now supports self-hosted machines for Cloud Agents. The September 2026 release keeps tool execution in infrastructure a team manages, while the agent handles the tool calls from Cursor.

Cursor logo
CursorFreemium

The AI-native code editor with $1B+ ARR, 25+ models, and background agents on dedicated VMs

Keep execution inside your network

Cursor says a self-hosted machine keeps the codebase, build outputs, and secrets on an internal machine rather than moving them into Cursor-hosted execution. That makes the feature relevant to teams with private services, custom hardware, or network-bound development environments.

The agent loop still needs a carefully configured connection to the worker. Self-hosting changes where tools run; it does not remove the need to review repository access, credentials, outbound network rules, and the commands an agent is allowed to execute.

Personal machines and dynamic team pools

My Machines connects one laptop or virtual machine to an account for personal workflows. Team pools provide named queues of workers for a team or enterprise. Cursor says pools can grow as requests arrive, shrink when machines disconnect, and assign work to any available worker rather than tying capacity to one repository.

Pools can also hibernate idle machines and restore them within a reconnect window when a follow-up arrives. That can reduce the need to keep expensive capacity warm, but teams should test startup and reconnect behavior before treating hibernation as an availability guarantee.

Run agents on existing sandboxes

Cloud Agents can also execute on infrastructure from supported providers including AWS Lambda, Coder, Cloudflare, Daytona, Modal, Namespace, Vercel, and E2B. The provider options give teams another way to place agent tool execution near existing development or deployment systems without moving every workflow onto a single machine.

Computer use on Linux and Mac

Self-hosted workers support computer use on Linux and Mac when the required desktop packages are installed. Cursor describes agents that can click, type, take screenshots, and drive a browser, with the user able to watch the desktop or take control from Cursor.

Computer-use access should be treated as a high-impact permission. Before enabling it, define the desktop account, browser profile, network boundary, and approval policy that apply to each worker.

Availability and evaluation checklist

The update is most useful for teams that need Cloud Agent workflows to reach internal systems or specialized hardware while keeping execution in their own environment. Evaluate worker registration, pool routing, reconnect behavior, secret handling, and computer-use permissions against a representative task before expanding access.

Sources

Free Resource

2026 AI Coding Tools Comparison Chart

Side-by-side comparison of features, pricing, and capabilities for every major AI coding tool.

No spam, unsubscribe anytime.

Frequently Asked Questions

What is Cursor brings Cloud Agent execution to self-hosted machines?
Cursor’s September 2026 self-hosted machines release keeps Cloud Agent tool execution, code, build outputs, and secrets inside a team’s own network, with dynamic worker pools and sandbox options.