A cloud development environment replaces "it works on my laptop" with one machine that has the right versions, that a new teammate can open in minutes, and that keeps running a long build after you close the lid. Setting one up comes down to four decisions:
- Size. CPU, memory and disk for your heaviest task: a full build, a test suite, an IDE's language server.
- Setup. How the repository, packages and services get onto the machine: a script, a Dockerfile, or a prepared image.
- Editor. VS Code in the browser, or a local editor over SSH.
- Idle time. What happens when nobody is working, and what it costs.
On Runtime, a dev environment is a Linux machine with VS Code in the
browser already running on port 8080 over /workspace. You size it, name it,
and reach it through a private HTTPS preview, and when nobody uses it, it
pauses with its memory kept: open terminals, running servers and the
editor's state come back as they were. (Dev environments)
Create it once, by name
getOrCreate with a name makes the environment the first time, and returns
the same one on every later call. This script creates it at 4 vCPU, 8 GB of
memory and 16 GB of disk, clones a repository, installs its packages, and
prints a link to VS Code:
TerminalAPI=https://api.withruntime.comAUTH="Authorization: Bearer $RUNTIME_API_KEY"JSON="Content-Type: application/json"ws=$(curl -sf -X POST "$API/v2/workspaces" -H "$AUTH" -H "$JSON" -d '{ "name": "my-dev", "getOrCreate": true, "vcpu": 4, "memoryGB": 8, "diskGB": 16, "idlePauseSeconds": 1800}')id=$(echo "$ws" | jq -r .id)# Put the project in /workspace and install what it needs.curl -sf -X POST "$API/v2/workspaces/${id}:exec" -H "$AUTH" -H "$JSON" -d '{ "cmd": "cd /workspace && { [ -d express ] || git clone --depth 1 https://github.com/expressjs/express; } && cd express && npm install --no-audit --no-fund"}' | jq -r .exitCode# Share VS Code's port, 8080, with a private preview and print the link.preview=$(curl -sf -X POST "$API/v2/sandboxes/${id}/previews" -H "$AUTH" -H "$JSON" \ -d '{"port": 8080}')echo "$(echo "$preview" | jq -r .url)?folder=/workspace/express&runtime_preview_token=$(echo "$preview" | jq -r .token)"Set the sizes in the request rather than relying on a default, so the
environment is the same for everyone who runs the script. Run it again
tomorrow and it returns the same machine, with the clone and node_modules
already there.
Keep a server running between sessions
Start long-running processes with "wait": false. The call returns at once
and the process keeps running:
Terminalcurl -s -X POST "$API/v2/workspaces/${id}:exec" -H "$AUTH" -H "$JSON" -d '{ "cmd": ["node", "-e", "require(\"http\").createServer((q, s) => s.end(\"hi\")).listen(3000)"], "wait": false}' | jq -r .state# runningA pause keeps that process in memory. After a pause and a resume, the same server still answers on port 3000, without being started again. To reach it from your browser, give port 3000 a preview of its own (share a port).
Connect a local editor instead
VS Code Remote - SSH and JetBrains Gateway can open the same kind of machine from your desktop. See use VS Code Remote - SSH and use JetBrains Gateway. For what a remote environment is and when it helps, see what is a remote development environment?.
Give every developer their own
Run the same script with a different name per person, my-dev-alice,
my-dev-bob, and each gets an isolated machine with its own kernel, disk and
network. Each one pauses when its owner stops working
(per-user dev environments).
Delete it
Terminalcurl -s -X DELETE "$API/v2/workspaces/${id}" -H "$AUTH" | jq -r .state# deletedDeleting removes the machine and its disk. Push your branches first.
Related
- How to open VS Code in the browser
- How to run Jupyter in the cloud
- What is a Jupyter kernel?
- What is a remote development environment?
- GitHub Codespaces vs an agent sandbox
Facts on this page were checked on 17 November 2026.