Runtm

Open-source sandboxes where coding agents build and deploy.

Spin up isolated environments where Claude Code, Cursor, Codex, and other agents code and ship software. With live URLs, logs, and previews.

Website · Docs · Get Started

Demo

https://github.com/user-attachments/assets/8d6d5ab8-a5c4-4a3d-8ef1-5d20b67ed3ee

Why Runtm?

• Sandboxes for yolo agents – Let agents code with full permissions in isolated environments. No risk to your machine or cloud.
• Run any coding agent – Claude Code, Cursor, Codex, Gemini CLI, and more. Bring your favorite.
• Real URLs instantly – Agents deploy to live HTTPS endpoints. Test, share, and iterate.
• Logs and observability – See what your agent built, debug issues, fix and redeploy.

Quick Start

# Install
uv tool install runtm

# Start a local sandbox session
runtm session start

# Your agent builds inside the sandbox...

# Deploy to a live URL
runtm session deploy

You get a live HTTPS endpoint on auto-stopping infrastructure. Machines spin down when idle and wake up on traffic.

Local Sandbox Sessions

Run AI agents in isolated local environments with OS-level sandboxing:

# Start a sandbox (auto-installs deps on first run)
runtm session start

# Start with a template
runtm session start --template web-app

# Use a different agent
runtm session start --agent codex

# List all sandboxes
runtm session list

# Reattach to a sandbox
runtm session attach sbx_abc123

# Deploy from sandbox to live URL
runtm session deploy

What you get:

• OS-level isolation – Uses bubblewrap (Linux) or seatbelt (macOS) for fast, secure sandboxing
• Instant startup – Sandboxes start in <100ms, no containers needed
• Multi-agent support – Works with Claude Code, Codex, Gemini CLI, and more
• Persistent workspaces – Stop and resume sessions, files preserved

How It Works

┌────────────┐     ┌─────────┐     ┌─────────┐     ┌──────────┐
│ Your Agent │ ──▶ │ Sandbox │ ──▶ │  Runtm  │ ──▶ │ Live URL │
└────────────┘     └─────────┘     └─────────┘     └──────────┘

1. Spin up a sandbox – Your agent gets an isolated workspace
2. Agent builds – Full permissions to code, install deps, run tests
3. Deploy – One command to a real URL with logs and previews
4. Iterate – Agent can see logs, fix bugs, and redeploy

Installation

# Recommended
uv tool install runtm

# Alternative
pipx install runtm

# Or with pip
pip install runtm

Commands

See the CLI docs for the full reference.

Self-Hosting

Runtm can be fully self-hosted. See the self-hosting guide.

git clone https://github.com/runtm-ai/runtm.git
cd runtm
cp infra/local.env.example .env

# Install packages (includes sandbox and agents)
./scripts/dev.sh setup

# Start local services
docker compose -f infra/docker-compose.yml up -d

# Use the development CLI
runtm-dev start                    # Start a sandbox session
runtm-dev prompt "Build an API"    # Send prompt to agent

Note: Use runtm-dev (not runtm) when self-hosting. The dev CLI includes sandbox/agents packages.

Project Structure

packages/
  shared/     # Types, manifest schema, errors
  sandbox/    # Local sandbox runtime (OS-level isolation)
  agents/     # AI coding agent adapters (Claude Code, Codex, etc.)
  api/        # FastAPI control plane
  worker/     # Build + deploy pipeline
  cli/        # Python CLI (Typer)

templates/    # Starter projects (backend, static, fullstack)

Documentation

Full docs at docs.runtm.com:

• Quickstart
• CLI Reference
• API Reference
• Self-Hosting

Contributing

See CONTRIBUTING.md for development setup and guidelines.

License

Support

Open an issue or join our Discord.