Runtm CLI
Runtm runs Claude Code, Codex and other agents in isolated sandboxes and deploys the app they build to a live HTTPS endpoint
- 303 ★ on the repository
- Last commit
Broad access
Why this level
- The agent works with full rights in the sandbox: installs packages and runs code and network calls inside the session
- `runtm session deploy` publishes the project at a live HTTPS address, so code and data are visible to anyone with the link
- Cloud features use an API key that lands in the environment of the agent and scripts
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/runtmDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The agent works with full rights in the sandbox: installs packages and runs code and network calls inside the session
- `runtm session deploy` publishes the project at a live HTTPS address, so code and data are visible to anyone with the link
- Cloud features use an API key that lands in the environment of the agent and scripts
This is third-party code. Review the repository files before installing.
What it does
Runtm is an open set of tools for running coding agents in a sandbox. `runtm session start` brings up a local session where the agent installs dependencies, writes code and runs tests freely. Isolation works at the OS level: bubblewrap on Linux and seatbelt on macOS, no containers needed. The README puts sandbox startup under 100 ms. The `--agent codex` flag picks another agent and `--template web-app` a project template. A stopped session keeps its files, and `runtm session attach <id>` returns to it. `runtm session deploy` publishes the finished app: it gets a live HTTPS endpoint on auto-stopping infrastructure, where a machine sleeps when idle and wakes on traffic. For a project outside a session there are `runtm init` and `runtm deploy`, and `runtm logs <id>`, `runtm status <id>` and `runtm destroy <id>` show logs, status and tear a deployment down. The agent reads the logs itself, fixes the error and redeploys. A separate Go CLI, `runtm-api`, is meant for agents that drive the hosted Cloud API: JSON in and out, stable exit codes and embedded skill files for Claude Code, Cursor and Codex. Both CLIs use one API key. You can self-host the server with `docker compose`, and then you call `runtm-dev` instead of `runtm`.
Who it is for. Developers who give Claude Code, Cursor or Codex full rights to build and test and want that to happen in a sandbox, with the result shown at a live link.
Good fit when
- An agent needs full rights to install dependencies and run tests, but not on your machine
- The result must be shown to colleagues at a live HTTPS address after `runtm session deploy`
- An agent should read deployment logs itself and redeploy a fix (`runtm logs`, `runtm status`)
- An agent drives the cloud API and needs JSON in and out with stable exit codes (`runtm-api`)
- Source code cannot go to a third-party host, so you deploy the server yourself with `docker compose`
Not a fit when
- You only need command isolation without deploying: a standalone sandbox such as Fence is enough
- The work machine runs Windows: the README mentions only bubblewrap for Linux and seatbelt for macOS
- The project must not reach the internet: a deployment publishes it at a live address
Example request
Run `runtm session start --template web-app`, build a simple FastAPI service in the sandbox and deploy it with `runtm session deploy`Limitations
- Deployments go to Runtm hosting, and both CLIs share one API key. Your own server comes up through `docker compose`, and then you call `runtm-dev`: the README advises it instead of `runtm` when self-hosting.
- The sandbox relies on bubblewrap on Linux and seatbelt on macOS, and the README is silent about Windows.
- Licenses differ: the server is AGPLv3, the CLI is Apache 2.0, the templates are MIT.
- The agent acts in the sandbox with full permissions, so it changes the project inside the session without limits.
How to disable. `runtm session stop <id>` stops a sandbox and keeps its files, `runtm session destroy <id>` deletes it with the files. `runtm destroy <id>` tears down a deployment. The CLI itself is removed with `uv tool uninstall runtm`.
FAQ
How do I install Runtm?
With `uv tool install runtm`, with `pipx install runtm` and `pip install runtm` as fallbacks. Then `runtm session start` launches a local sandbox and `runtm session deploy` publishes the project at a live address.
How does `runtm` differ from `runtm-api`?
`runtm` is written in Python and covers local sandboxes, templates and deploys. `runtm-api` is written in Go, lighter, and built for agents that drive the hosted Cloud API. It installs with `curl -fsSL https://runtm.com/install | bash`, and both share one key.
Which agents run in a Runtm sandbox?
Claude Code, Codex, Gemini CLI and others: the `--agent` flag picks one, for example `--agent codex`. The README names Cursor as well.
What is the isolation made of?
OS facilities: bubblewrap on Linux and seatbelt on macOS. The local sandbox needs no containers.
Can I self-host Runtm?
Yes. Clone the repository, copy `infra/local.env.example` to `.env`, run `./scripts/dev.sh setup` and `docker compose -f infra/docker-compose.yml up -d`, then call `runtm-dev`.
Discussion
No questions or reviews yet.
Related
CLI toolsAgentBox runs Claude Code, Codex and Open Code in separate virtual machines on your PC, a self-hosted server or a cloud.
A Rust sandbox for parallel coding agents: each gets its own workspace over one shared project, and changes publish as one set.
A self-hosted control center for coding agents: Claude Code, Codex, Copilot, OpenCode and Pi from your phone, desktop and terminal
Containarium: isolated boxes for AI agents
Containarium
A platform on your own server that gives each agent its own Linux box with SSH and MCP tools for commands and files
A Docker sandbox for coding agents: a Debian image with Copilot CLI, Codex and Pi, and SSH, RDP and Docker enabled by flags.
The vp run command starts Claude Code, Codex, Gemini and other CLI agents in Docker or Podman containers
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Runtm (this card) | yes | 303 | |
| Fence | yes | 1,012 | runs commands and agents in a sandbox with the network blocked by default, but does not deploy to a live address |
| Vercel Sandbox | no | 219 | Vercel's official skill for isolated micro-VMs running agent code, while Runtm ties the sandbox to its own deployment |
README in short
The README states the idea in one line and explains why sandboxes matter. The quick start takes four commands, then come local sessions with a description of the isolation and an agent, sandbox, Runtm, live URL diagram. Then it covers installing the two CLIs, a command table, self-hosting and the package structure.