runtm 0.2.19

Sandboxes where AI agents build and deploy. Spin up isolated environments for Claude Code, Cursor, and other agents.

runtm

CLI for Runtm – sandboxes where AI agents build and deploy.

Website: runtm.com · Docs: docs.runtm.com · Sign up: app.runtm.com

Installation

Recommended (uv):

uv tool install runtm

Alternative (pipx):

pipx install runtm

From PyPI (pip):

pip install runtm

With Sandbox Support

To use local sandboxes with AI agents:

pip install runtm[sandbox]

Upgrading

# Upgrade to latest version (uv)
uv tool upgrade runtm

# Or force reinstall
uv tool install runtm --force

# With pipx
pipx upgrade runtm

# With pip
pip install --upgrade runtm

Quick Start

# 1. Authenticate with Runtm
runtm login

# 2. Start a sandbox and build with AI
runtm start
runtm prompt "Build a REST API with SQLite"

# 3. Deploy to a live URL
runtm deploy

You get a live HTTPS endpoint. Machines auto-stop when idle and wake on traffic.

Commands

Sandbox Commands

Command	Description
runtm start	Start a sandbox session (interactive menu)
runtm prompt "..."	Send a prompt to the agent (autopilot mode)
runtm attach [id]	Attach to a sandbox (defaults to active)
runtm session list	List all sandbox sessions
runtm session stop <id>	Stop a session (preserves workspace)
runtm session destroy <id>	Destroy a session and delete workspace
runtm session deploy	Deploy from sandbox to live URL

Project Commands

Command	Description
runtm init [template]	Initialize from template (backend-service, web-app, static-site)
runtm run	Run project locally (auto-detects runtime)
runtm validate	Validate project before deployment
runtm fix	Auto-fix common issues (lockfiles)
runtm deploy [path]	Deploy project to a live URL

Deployment Commands

Command	Description
runtm status <id>	Show deployment status
runtm logs <id>	Show logs (build, deploy, runtime)
runtm list	List all deployments
runtm search <query>	Search deployments by description/tags
runtm destroy <id>	Destroy a deployment

Configuration Commands

Command	Description
runtm config set/get/list	Manage CLI configuration
runtm secrets set/get/list/unset	Manage environment secrets
runtm domain add/status/remove	Manage custom domains
runtm approve	Apply agent-proposed changes

Authentication Commands

Command	Description
runtm login	Authenticate with Runtm API
runtm logout	Remove saved credentials
runtm doctor	Check CLI setup and diagnose issues
runtm version	Show CLI version

Sandbox Sessions

Start isolated environments where AI agents can build software:

# Start with interactive menu
runtm start

# Or go directly to autopilot mode
runtm start --autopilot

# Send prompts to the agent
runtm prompt "Build a todo API with SQLite"
runtm prompt --continue "Add authentication"

# Attach to see what's happening
runtm attach

# List all sessions
runtm session list

Modes

• Autopilot: Agent runs autonomously, control via runtm prompt
• Interactive: Drop into sandbox shell, control agent manually

Available Agents

• claude-code - Anthropic's Claude Code (recommended)
• codex - OpenAI's Codex CLI
• gemini - Google's Gemini CLI

Authentication

Get your free API key at app.runtm.com. The CLI will prompt you to authenticate on first use.

# Manual login
runtm login

# Login with token directly
runtm login --token runtm_sk_xxx

# Check auth status
runtm doctor

# Logout
runtm logout

Token storage:

• Primary: ~/.runtm/credentials file (0o600 permissions)
• Optional: System keychain (if keyring package installed)

Environment variable override:

export RUNTM_API_KEY=runtm_sk_xxx  # Overrides stored token

Configuration

# Set API URL (for self-hosting)
runtm config set api_url=https://self-hosted.example.com/api

# Get a config value
runtm config get api_url

# List all config values
runtm config list

# Reset to defaults
runtm config reset

Config file: ~/.runtm/config.yaml

Environment variables:

• RUNTM_API_URL - API endpoint (overrides config)
• RUNTM_API_KEY - API key (overrides stored token)
• RUNTM_DEBUG - Enable debug logging

Secrets Management

Manage environment variables for deployments:

# Set secrets
runtm secrets set DATABASE_URL=postgres://...
runtm secrets set API_KEY=sk-xxx

# List secrets
runtm secrets list

# Get a secret value
runtm secrets get DATABASE_URL

# Remove a secret
runtm secrets unset OLD_KEY

Secrets are stored in .env.local (gitignored) and injected at deploy time.

Troubleshooting

# Check CLI setup and diagnose issues
runtm doctor

Example output:

runtm v0.2.7
  API URL:      https://app.runtm.com/api
  Auth storage: keychain ([e-mail])
  Auth status:  ✓ Authenticated as [e-mail]
  Connectivity: ✓ API reachable (142ms)

  Ready to deploy! Run: runtm init

Machine Tiers

All deployments use auto-stop for cost savings (machines stop when idle and start automatically on traffic).

Tier	CPUs	Memory	Est. Cost	Use Case
starter (default)	1 shared	256MB	~$2/month*	Simple tools, APIs
standard	1 shared	512MB	~$5/month*	Most workloads
performance	2 shared	1GB	~$10/month*	Full-stack apps

*Costs are estimates for 24/7 operation. With auto-stop, costs are much lower for low-traffic services.

Deployment

# Deploy to a live URL (uses starter tier by default)
runtm deploy

# Deploy with a specific tier
runtm deploy --tier standard
runtm deploy --tier performance

# Check deployment status
runtm status dep_abc123

# View logs
runtm logs dep_abc123

Redeployment (CI/CD)

Runtm supports automatic redeployment based on the project name in runtm.yaml:

# First deploy - creates new deployment
runtm deploy                   # → v1, creates new URL

# Fix a bug, then redeploy - updates existing
runtm deploy                   # → v2, same URL, updated code

# Force a completely new deployment
runtm deploy --new             # → v1, new deployment, new URL

Logs

# All logs (build + deploy + recent runtime)
runtm logs dep_abc123

# Filter by log type
runtm logs dep_abc123 --type runtime
runtm logs dep_abc123 --type build

# More runtime log lines
runtm logs dep_abc123 --lines 100

# Search logs
runtm logs dep_abc123 --search "error"
runtm logs dep_abc123 --search "error,warning,timeout"  # OR logic

# Pipe to grep (Heroku-style)
runtm logs dep_abc123 --raw | grep "error"

# JSON output for AI agents
runtm logs dep_abc123 --json

Development

# Install in editable mode with sandbox support
pip install -e ".[dev,sandbox]"
pip install -e ../sandbox
pip install -e ../agents

# Use the development CLI (avoids conflicts with PyPI version)
runtm-dev start                    # Start sandbox session
runtm-dev prompt "Build an API"    # Send prompt to agent
runtm-dev session list             # List sessions

# Configure CLI to use local API (add to ~/.zshrc or ~/.bashrc)
export RUNTM_API_URL=http://localhost:8000
export RUNTM_API_KEY=dev-token-change-in-production

# Run tests
pytest

runtm vs runtm-dev

CLI	Source	Use Case
runtm	PyPI (pip install runtm)	Production use
runtm-dev	Local .venv/	Development (includes sandbox/agents)

If you have the PyPI version installed globally, use runtm-dev to ensure you're running your local development code with full sandbox support.