Skip to the content
compare

Same work. Different place to run it.

Your laptop keeps the setup close. Hosted platforms hand operation to a provider. Cloudfield moves agent sessions to a Linux server you operate, where your whole team can watch and steer them. Choose the responsibilities you want to keep.

tradeoffs

What changes, side by side

Here, “your laptop” means an agent running locally. Hosted platforms are a category, not one product: check a provider's current documentation before choosing.

Execution, collaboration and operating responsibilities
What changesYour laptopHosted platformsCloudfield
Execution hostYour computer runs the agent.Provider-operated compute. Available environments vary.A Linux server you operate runs the agent in Docker.
Laptop sleepLocal execution pauses when the computer sleeps.Remote execution need not depend on your laptop. Session limits vary.Disconnect without stopping the session. Your server must stay on.
Agent choiceUse agents that support your local environment.Agent and model choices depend on the provider.Claude Code, Codex and many more agent CLIs, or a custom agent of your own, chosen per session. Cloudfield installs one when you ask and you log in on the server.
Working togetherGit and collaboration tools you choose.Live access, permissions and review workflows vary.Everyone with access to a workspace can watch its live sessions. One person controls a session at a time and a teammate can ask to take over. Viewer, collaborator, admin and super admin roles set who may steer.
Environment / isolationYour local environment, plus any sandbox you configure.Sandbox boundaries and configuration vary by provider.A container and git worktree per session, plus its own headless browser once opened. The container is the boundary; agents run unprivileged, and a permission tier sets what sudo may do.
Returned codeAlready in your local checkout; review and commit with your tools.Branches, pull requests and export options vary.Every session commits to its own branch. Push it and open a pull request from the app, or pull it into your clone; neither merges it into your base branch.
Accounts / credentials / dataLocal credentials and the data policies of your chosen agent and model provider.Account requirements, credential handling and data retention vary. Check the provider's terms.Credentials live in server-side member homes mounted into sessions. Agent providers still receive the data their CLIs send. Optional edge details below.
Operator / costYou maintain the computer and pay your chosen agent or model provider.The provider operates the service. Plans, limits and included model usage vary.You maintain the server, Docker, updates and backups, and cover compute plus agent or model costs.

account sharing

Account sharing includes credentials

A session on a shared account belongs to the launcher and runs in the launcher's environment. The owner's account supplies the agent's login file, read-write, plus quota and cost attribution. One agent CLI is an exception and shares the owner's whole agent directory; the security guide names it. Revoking a share blocks future launches and reopens and moves running sessions to another account the launcher may use, or stops them. It does not recall a login a session copied. Read the security boundary.

edge metadata

A relay still keeps metadata

The optional publisher-operated edge stores identity and connection metadata, but cannot read end-to-end encrypted session content. That is not the same as no data reaching the publisher. Read the privacy guide.

hosting

The workspace stays with you

There is no hosted Cloudfield workspace. The optional edge is available today as a sign-in and relay service; your server still runs the agents. Read about the edge.

fit

Three reasonable choices

local

Keep the work local

  • You want the shortest path to your existing local tools, and your work fits on one machine.
  • Keeping it awake is acceptable; you do not need a shared, always-on place for agent sessions.

hosted

Let a provider operate

  • You prefer a provider-operated service whose agents, integrations and data terms suit your work.
  • Its plan and limits make more sense than maintaining a server.

self-hosted

Run Cloudfield on your server

  • You are a startup or small team using coding agents, and you want to watch and steer each other's sessions live, not only read the result.
  • You have a Linux server that can stay on and someone to maintain it, and you review agent work as branches and pull requests.

Cloudfield is the third option. Your hardware, your code.