A TUI + CLI for the Runloop.ai platform. Use it as an interactive TUI (Terminal User Interface) with rich UI components, or as a traditional CLI for scripting and automation.
# TUI mode - launches an interactive terminal UI rli # CLI mode - perfect for scripts and automation rli devbox list # Outputs JSON/text rli devbox create --name my-devbox rli devbox exec <devbox-id> echo "Hello World" rli devbox delete <devbox-id>
- 🖥️ TUI mode — Interactive terminal UI with menus, tables, and real-time updates
- 🎯 CLI mode — Traditional commands with text, JSON, and YAML output for scripting
- ⚡ Fast and responsive with pagination
- 📦 Manage devboxes, snapshots, and blueprints
- 🚀 Execute commands, PTY shell, view logs in devboxes (SSH access was retired; use
rli devbox ptyfor an interactive shell) - 🤖 Model Context Protocol (MCP) server for AI integration
Install globally via npm or pnpm:
npm install -g @runloop/rl-cli
# or
pnpm add -g @runloop/rl-cliexport RUNLOOP_API_KEY=your_api_key_hereGet your API key from https://runloop.ai/settings
By default the CLI and MCP server connect to https://api.runloop.ai. To use a different deployment, set RUNLOOP_BASE_URL to the full API URL:
export RUNLOOP_BASE_URL=https://api.runloop.proThe URL must be of the form https://api.<domain>, optionally with a :port (e.g. https://api.runloop.pro:8443). The CLI derives other service hostnames from the domain portion:
| Service | Host |
|---|---|
| API | https://api.<domain> (the value of RUNLOOP_BASE_URL) |
| Platform | https://platform.<domain> |
| Tunnels | tunnel.<domain> (PTY sessions reach the devbox over a tunnel created via the API) |
rli # Launch the interactive TUI rli --help # See help information
All commands support --output (-o) for format control:
rli devbox list # Default text output rli devbox list -o json # JSON output rli devbox list -o yaml # YAML output
The CLI is organized into command buckets:
rli devbox create # Create a new devbox rli devbox list # List all devboxes rli devbox delete <id> # Shutdown a devbox rli devbox exec <id> <command...> # Execute a command in a devbox rli devbox exec-async <id> <command...> # Execute a command asynchronously on a... rli devbox upload <id> <file> # Upload a file to a devbox rli devbox get <id> # Get devbox details rli devbox get-async <id> <execution-id> # Get status of an async execution rli devbox suspend <id> # Suspend a devbox rli devbox resume <id> # Resume a suspended devbox rli devbox shutdown <id> # Shutdown a devbox rli devbox pty <id> # Connect to a devbox PTY session via W... rli devbox read <id> # Read a file from a devbox using the API rli devbox write <id> # Write a file to a devbox using the API rli devbox download <id> # Download a file from a devbox rli devbox send-stdin <id> <execution-id> # Send stdin to a running async execution rli devbox logs <id> # View devbox logs
rli snapshot list # List all snapshots rli snapshot create <devbox-id> # Create a snapshot of a devbox rli snapshot delete <id> # Delete a snapshot rli snapshot get <id> # Get snapshot details rli snapshot prune <devbox-id> # Delete old snapshots for a devbox, ke... rli snapshot status <snapshot-id> # Get snapshot operation status
rli blueprint list # List all blueprints rli blueprint create # Create a new blueprint rli blueprint get <name-or-id> # Get blueprint details by name or ID (... rli blueprint logs <name-or-id> # Get blueprint build logs by name or I... rli blueprint delete <id> # Delete a blueprint by ID rli blueprint prune <name> # Delete old blueprint builds, keeping ... rli blueprint from-dockerfile # Create a blueprint from a Dockerfile ...
rli object list # List objects rli object get <id> # Get object details rli object download <id> [path] # Download an object. Omit path to save... rli object upload [paths...] # Upload an object. Reads from piped st... rli object delete <id> # Delete an object (irreversible)
rli network-policy list # List network policies rli network-policy get <id> # Get network policy details rli network-policy create # Create a new network policy rli network-policy delete <id> # Delete a network policy
rli secret create <name> # Create a new secret. Value can be pip... rli secret list # List all secrets rli secret get <name> # Get secret metadata by name rli secret update <name> # Update a secret value (value from std... rli secret delete <name> # Delete a secret
rli gateway-config list # List gateway configurations rli gateway-config create # Create a new gateway configuration rli gateway-config get <id> # Get gateway configuration details rli gateway-config update <id> # Update a gateway configuration rli gateway-config delete <id> # Delete a gateway configuration
rli mcp-config list # List MCP configurations rli mcp-config create # Create a new MCP configuration rli mcp-config get <id> # Get MCP configuration details rli mcp-config update <id> # Update an MCP configuration rli mcp-config delete <id> # Delete an MCP configuration
rli mcp start # Start the MCP server rli mcp install # Install Runloop MCP server configurat...
rli axon list # List active axons rli axon events <id> # List events for an axon
rli agent list # List agents rli agent create # Create a new agent rli agent delete <id-or-name> # Delete an agent rli agent show <id-or-name> # Show agent details
Runloop includes a Model Context Protocol (MCP) server that allows AI assistants like Claude to interact with your devboxes.
# Install MCP configuration rli mcp install # Restart Claude Desktop, then ask Claude: # "List my devboxes" or "Create a new devbox"
# Stdio mode (for Claude Desktop) rli mcp start # HTTP mode (for web/remote access) rli mcp start --http rli mcp start --http --port 8080
Documentation:
- CLAUDE_SETUP.md - Complete setup guide for Claude Desktop
- MCP_README.md - Full MCP documentation
- MCP_COMMANDS.md - Quick command reference
The TUI supports both light and dark terminal themes and will automatically select the appropriate theme.
# Install dependencies pnpm install # Build pnpm run build # Watch mode pnpm run dev
If the TUI crashes (e.g. when pressing Enter on a form field), you can capture logs to inspect the error:
Option 1 – stderr to file (no debug env)
Useful to see uncaught errors and stack traces that the app writes to stderr:
pnpm run build
pnpm run start:debug
# Reproduce the crash, then:
cat debug.logOption 2 – run under Node with inspector
To get a stack trace from an uncaught exception, run with Node’s inspector and reproduce the crash; the process will pause and you can inspect the stack:
node --inspect-brk dist/cli.js
# Attach Chrome/Edge to the URL shown (e.g. chrome://inspect) and resume; reproduce the crash.We welcome contributions! Please see CONTRIBUTING.md for guidelines on how to contribute to this project.
MIT