Your coding agents keep working after you close the laptop.
Hand off the task. Close the laptop. Check in from your phone.
Already run Claude Code, Codex or Gemini CLI on an API key? Ask to join the existing-API-key cohort →
Your agent is only as useful as the machine it runs on. Today, that's your laptop.
- 01Close the lid and the work stops.
- 02One checkout means one task at a time, or agents tripping over each other's changes.
- 03Every new task starts with cloning, installing and waiting.
- 04You can't see or steer it unless you're at that laptop.
- 05The agent holds your keys, in a shell it controls.
Close the laptop. The work carries on.
- The workspace runs somewhere else, so closing the tab or the lid stops nothing.
- Send a message and walk away. If the workspace is still starting, the message waits and is delivered when it's up.
- A paused workspace keeps its disk, and picking it up again carries on the same conversation.
One task, one workspace. As many as you have tasks.
- Each task gets its own workspace: its own checkout, disk and shell.
- Two tasks never share a working tree.
- Start as many as you have tasks. Your plan sets how many run at once, and the rest wait their turn and start on their own.
Start where you left off, not from a fresh clone.
- Start from a template for your stack (Ubuntu, Node, Python or Rust) or write your own.
- An optimized template has already done the setup. A new workspace starts from a saved copy of the disk.
- Every plan can optimize templates on your own Docker account or machine.
The whole app, on your phone.
- Add it to your home screen.
- Read the transcript, answer the agent's questions, steer it or queue the next message, and start a new task.
On your Docker account, your own machine, or your own cluster.
- Docker Sandboxes: run on your own Docker account, connected from the app.
- Your own machine: free, on a Linux machine you already have.
- Your own cluster (Enterprise): workspaces and checkouts stay in your Kubernetes cluster.
Your agent, your plan.
Bring Claude Code, Codex or Gemini. Claude Code and Codex run on the plan you already have; Gemini runs on your API key.
smedja never bills you for tokens.
Pricing →Assistants that react so you don't.
Mention one in Slack, delegate a Linear ticket, have one review each pull request, or put one on a schedule.
Integrations →Your keys never enter the workspace.
Model keys go through a proxy, git tokens are filled in per operation, and MCP credentials are held by smedja.
Security →It works with what you have.
GitHub and gitlab.com, private and public, to clone and push. Public repositories on any other git host. Slack, Linear, CircleCI and any remote MCP server.
Integrations →Scriptable, too.
Start workspaces, hand off tasks and read what the agent did from the terminal or over MCP, as well as in the app. An agent can start the next workspace itself.
Quickstart →