How Do You Run Coding Agents in the Cloud? Cline Cloud Explained

Hey hey hey, welcome to Agentic Pod! I'm Liam.

And I'm Stella! Liam, I closed my laptop last night with a coding task half done. Is it true Cline can now keep working without me?

It is. Cline Cloud runs up to ten coding agents at the same time, each in its own cloud sandbox. Laptop shut, Wi-Fi off, they keep going.

Ten agents. I can barely manage one tab. Okay, how do I try it?

Start simple. What is it, exactly?

It's Cline's coding agent, but it runs on Cline's servers instead of your machine. You control it from your browser. Nothing to install.

Not even an editor?

No editor, no CLI. The release notes from October 16th say it's rolling out gradually, and cloud tasks show up in Recent and History with a status. Cline notifies you when one finishes.

So where does the code go?

To GitHub. Each session works on its own branch that starts with cline slash, and it commits and pushes as it goes. So you can watch progress on GitHub before it's even done.

And it can't touch my main branch.

Right. And the GitHub token for each session is limited to the one repo you picked.

Okay, show me the steps.

Five steps. One, go to app.cline.bot and sign in. It's a free account. Two, connect GitHub and choose which repos Cline can see.

Personal repos and work ones?

Both, the ones you belong to. Three, open the Agents tab and pick a repo and a branch. Four, pick a model, and describe the task. You can drag in a screenshot or a mockup.

Ooh, so I can paste a design and say "build this."

Yep. Five, optionally start in Plan mode, so Cline agrees on the approach with you before it writes any code. When it's done, ask it to open a pull request, then review and merge on GitHub.

And money? Is this one of those "free until it isn't" things?

There are three options. ClineFree has free models, though the list changes. ClinePass is a subscription for open-weight models, with two to five times the usage of standard API rates. And pay as you go with Cline credits for frontier models.

And what do I actually ask it to do?

Here's a real way to use the parallel part. Say you have four small jobs. Fix the flaky login test. Add dark mode to the settings page. Write docs for the export function. Upgrade one dependency.

Four sessions, four branches.

Exactly. Start all four, each with a clear prompt, and each one gets its own sandbox and its own copy of the repo. They can't step on each other. You get four pull requests to review over coffee.

That's way better than doing them one by one.

And the sidebar shows every session's status. You can send a follow-up to redirect any of them, or switch models mid-session if one is stuck. Prompt tip: say what done looks like. "Make the test pass and don't touch anything else."

And you can do this from your phone?

The cloud agents page works in a mobile browser. Check progress, answer an approval, start a session from the couch.

Okay, I need the catch.

Four. One, local rules, skills, MCP servers and connectors from your local Cline are not loaded in cloud sessions. So a setup you tuned on your laptop won't follow you.

Ouch. That's a big one.

Two, limits. Ten sessions at once, and fifty new sessions per day. Three, an idle session pauses after fifteen minutes and resumes with files intact. And four, it's new, so expect rough edges. There's a GitHub issue about a finished cloud task still showing as running.

So who's it for?

Developers with a backlog of small, well-defined jobs, people who code on the go, and anyone who wants to try an agent without installing anything. For big risky refactors, keep your hands on the wheel.

My verdict. Start with one boring task and a free model, and see what comes back.

Same. Links to the docs and the guide are in the description. Which chore would you hand to a cloud agent first? Tell us in the comments. See you next time!